divexplorer_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "db905f0a",
   "metadata": {},
   "source": [
    "# Environment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "powered-information",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-04-20T15:20:44.908897Z",
     "start_time": "2022-04-20T15:20:44.883774Z"
    },
    "id": "powered-information"
   },
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "built-doctor",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-04-20T15:20:45.349495Z",
     "start_time": "2022-04-20T15:20:44.910825Z"
    },
    "id": "built-doctor"
   },
   "outputs": [],
   "source": [
    "import divexplorer \n",
    "import pandas as pd\n",
    "pd.set_option('max_colwidth', None)\n",
    "import os\n",
    "import numpy as np\n",
    "\n",
    "from utils_analysis import filter_itemset_df_by_attributes, slice_by_itemset, plot_true_pred, plotComparisonShapleyValues, plotMultipleSV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "identical-thirty",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-04-20T15:20:45.366120Z",
     "start_time": "2022-04-20T15:20:45.351832Z"
    },
    "id": "identical-thirty"
   },
   "outputs": [],
   "source": [
    "## Define the minimum support threshold for data subgroups\n",
    "min_sup = 0.03"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "5e7e0b10",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Define abbreviations for plot and visualization\n",
    "from divexplorer.FP_Divergence import abbreviateDict\n",
    "abbreviations = {'Self-reported fluency level=native': 'fluency=native', \\\n",
    "                  'total_silence':'tot_silence', 'location': 'loc', \\\n",
    "                  'Current language used for work/school=English (United States)': 'lang=EN_US', \\\n",
    "                  'speakerId' : 'spkID', \\\n",
    "                  'First Language spoken=English (United States)':  'lang=EN_US', \\\n",
    "                  'trimmed':'trim', \\\n",
    "                  'total_':'', \\\n",
    "                  'speed_rate_word':'speakRate', \\\n",
    "                  'speed_rate_char':'speakCharRate', \\\n",
    "                  'change language': 'change lang', \\\n",
    "                  'duration': 'dur'}\n",
    "\n",
    "abbreviations_shorter = abbreviations.copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c80058f0",
   "metadata": {},
   "source": [
    "# Util Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "1fcc1cbb",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Function for computing the percentage of error for the problematic speaker in a given slice\n",
    "def percentage_problematic_speaker(df_sel, max_perc = 0.25, column = \"prediction\", groupby = \"speakerId\", problematic_speaker = \"NgQEvO2x7Vh3xy2xz\"):\n",
    "    \"\"\" Check if source of error is for more that max_perc of the problematic speaker\n",
    "        df_sel (pd.DataFrame) : dataset\n",
    "        max_perc (float): max percentage.\n",
    "    \n",
    "        Returns bool, percentage_error: True if percentage of error of the problematic speaker in the slice is greater than threshold\n",
    "    \"\"\"\n",
    "    df_sel_sk = df_sel.copy()\n",
    "    if \"speakerId\" not in df_sel_sk.columns:\n",
    "        df_sel_sk['speakerId'] = df_sel_sk.index.map(lambda x: x.split(\"/\")[3])\n",
    "    \n",
    "    tot_prediction_count = dict(df_sel_sk[column].value_counts())\n",
    "    if 0 not in tot_prediction_count:\n",
    "        return False, 0\n",
    "    tot_error_slice = tot_prediction_count[0]\n",
    "    df_sel_i = df_sel_sk.loc[df_sel_sk[groupby] == problematic_speaker]\n",
    "    if df_sel_i.shape[0]==0:\n",
    "        return False, 0\n",
    "    prediction_counts = dict(df_sel_i[column].value_counts())\n",
    "    if 0 not in prediction_counts:\n",
    "        return False, 0\n",
    "    percentage_error = prediction_counts[0]/tot_error_slice\n",
    "    return percentage_error > max_perc, percentage_error\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "c26247e4",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Function for sorting data cohorts\n",
    "def sortItemset(x, abbreviations={}):\n",
    "    x = list(x)\n",
    "    x.sort()\n",
    "    x = \", \".join(x)\n",
    "    for k, v in abbreviations.items():\n",
    "        x = x.replace(k, v)\n",
    "    return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "e539fc6e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def attributes_in_itemset(itemset, attributes, alls = True):\n",
    "    \"\"\" Check if attributes are in the itemset (all or at least one)\n",
    "    \n",
    "    Args:\n",
    "        itemset (frozenset): the itemset\n",
    "        attributes (list): list of itemset of interest\n",
    "        alls (bool): If True, check if ALL attributes of the itemset are the input attributes. \n",
    "        If False, check AT LEAST one attribute of the itemset is in the input attributes.\n",
    "        \n",
    "    \"\"\"\n",
    "    # Avoid returning the empty itemset (i.e., info of entire dataset)\n",
    "    if itemset == frozenset() and attributes:\n",
    "        return False\n",
    "    \n",
    "    for item in itemset:\n",
    "        # Get the attribute\n",
    "        attr_i = item.split(\"=\")[0]\n",
    "        \n",
    "        #If True, check if ALL attributes of the itemset are the input attributes.\n",
    "        if alls:\n",
    "            # Check if the attribute is present. If not, the itemset is not admitted\n",
    "            if attr_i not in attributes:\n",
    "                return False\n",
    "        else:\n",
    "            # Check if least one attribute. If yes, return True\n",
    "            if attr_i in attributes:\n",
    "                return True\n",
    "    if alls:\n",
    "        # All attributes of the itemset are indeed admitted\n",
    "        return True\n",
    "    else:\n",
    "        # Otherwise, it means that we find None\n",
    "        return False\n",
    "    \n",
    "def filter_itemset_df_by_attributes(df: pd.DataFrame, attributes: list, alls = True, itemset_col_name: str = \"itemsets\") -> pd.DataFrame:\n",
    "    \"\"\"Get the set of itemsets that have the attributes in the input list (all or at least one)\n",
    "    \n",
    "    Args:\n",
    "        df (pd.DataFrame): the input itemsets (with their info). \n",
    "        attributes (list): list of itemset of interest\n",
    "        alls (bool): If True, check if ALL attributes of the itemset are the input attributes. \n",
    "        If False, check AT LEAST one attribute of the itemset is in the input attributes.\n",
    "        itemset_col_name (str) : the name of the itemset column, \"itemsets\" as default\n",
    "        \n",
    "    Returns:\n",
    "        pd.DataFrame: the set of itemsets (with their info)\n",
    "    \"\"\"\n",
    "\n",
    "    return df.loc[df[itemset_col_name].apply(lambda x: attributes_in_itemset(x, attributes, alls = alls))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "7ba23a3a",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Function for substituting gender of a given itemset\n",
    "substitute_gender = lambda x : \"gender=male\" if \"gender=female\" in x else (\"gender=female\" if \"gender=male\" in x else x)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "occupational-madrid",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-04-20T15:07:23.652910Z",
     "start_time": "2022-04-20T15:07:23.612488Z"
    },
    "id": "occupational-madrid"
   },
   "source": [
    "# Define targets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "461193a2",
   "metadata": {
    "id": "461193a2"
   },
   "outputs": [],
   "source": [
    "## Target for DivExplorer: \n",
    "# 'prediction' is 1 if predicted_intet == original_intent, 0 otherwise\n",
    "target_col = 'prediction' \n",
    "target_metric = 'd_posr'\n",
    "target_div = 'd_accuracy'\n",
    "t_value_col = 't_value_tp_fn'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "8efa3b44",
   "metadata": {
    "id": "8efa3b44"
   },
   "outputs": [],
   "source": [
    "## Columns for visualization\n",
    "show_cols = ['support', 'itemsets', '#errors', '#corrects', 'accuracy', \\\n",
    "                'd_accuracy', 't_value', 'support_count', 'length']\n",
    "remapped_cols = {'tn': '#errors', 'tp': '#corrects', 'posr': 'accuracy', \\\n",
    "                target_metric: target_div, 't_value_tp_fn': 't_value'}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "young-mayor",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-04-20T15:26:21.281884Z",
     "start_time": "2022-04-20T15:26:21.260270Z"
    },
    "id": "young-mayor"
   },
   "outputs": [],
   "source": [
    "## Columns of the df file that we are going to analyze \n",
    "demo_cols = ['Self-reported fluency level ', 'First Language spoken',\n",
    "       'Current language used for work/school', 'gender', 'ageRange']\n",
    "\n",
    "slot_cols = ['action', 'object', 'location']\n",
    "\n",
    "signal_cols = ['total_silence', 'total_duration', 'trimmed_duration', \n",
    "       'n_words', 'speed_rate_word', 'speed_rate_word_trimmed'] \n",
    "\n",
    "input_cols = demo_cols + slot_cols + signal_cols "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "302b8cee",
   "metadata": {},
   "source": [
    "# Retrieve Data and Compute Divergence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "a9f27c91",
   "metadata": {},
   "outputs": [],
   "source": [
    "from divexplorer.FP_DivergenceExplorer import FP_DivergenceExplorer\n",
    "from divexplorer.FP_Divergence import FP_Divergence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "164157ed",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "wav2vec2-base\n",
      "wav2vec2-large\n",
      "hubert-base\n",
      "hubert-large\n"
     ]
    }
   ],
   "source": [
    "FP_fm_dict = {}\n",
    "fp_divergence_dict = {}\n",
    "df_dict = {}\n",
    "\n",
    "for config in [\"wav2vec2-base\", \"wav2vec2-large\", \"hubert-base\", \"hubert-large\"]:\n",
    "\n",
    "    print(config)\n",
    "\n",
    "    ## Read csv file\n",
    "    input_file_divexplorer = os.path.join(\\\n",
    "            os.getcwd(), \"data_precomputed\", f\"FSC_for_DivExplorer_{config}.csv\") \n",
    "\n",
    "    df = pd.read_csv(input_file_divexplorer, index_col = 0 )\n",
    "    df[target_col] = df[target_col]\n",
    "\n",
    "    ## Add SpeakerID information if it is present in the df\n",
    "    if \"speakerId\" in input_cols:\n",
    "        df['speakerId'] = df.index.map(lambda x: x.split(\"/\")[2])\n",
    "\n",
    "    ## Discretize the dataframe\n",
    "    from util_discretization import discretize\n",
    "\n",
    "    df_discretized = discretize(\n",
    "        df[input_cols+[target_col]],\n",
    "        bins=3,\n",
    "        attributes=input_cols,\n",
    "        strategy=\"quantile\", \n",
    "        round_v = 2,\n",
    "        min_distinct=5,\n",
    "    )\n",
    "\n",
    "    ## Replace values with ranges: \"low\", \"medium\", \"high\"\n",
    "    replace_values = {}\n",
    "\n",
    "    for i in range(0,len(signal_cols)):\n",
    "\n",
    "        for v in df_discretized[signal_cols[i]].unique():\n",
    "            if \"<=\" == v[0:2]:\n",
    "                replace_values[v] = \"low\"\n",
    "            elif \">\" == v[0]:\n",
    "                replace_values[v] = \"high\"\n",
    "            elif \"(\"  == v[0] and \"]\"  == v[-1]:\n",
    "                replace_values[v] = \"medium\"\n",
    "            else:\n",
    "                raise ValueError(v)\n",
    "\n",
    "        df_discretized[signal_cols[i]].replace(replace_values, inplace=True)\n",
    "        \n",
    "    df_discretized.loc[df_discretized[\"location\"]==\"none_location\", \"location\"] = \"none\"\n",
    "    df_discretized.loc[df_discretized[\"object\"]==\"none_object\", \"object\"] = \"none\"\n",
    "    \n",
    "    ## Create dict of Divergence df\n",
    "    df_dict[config] = df_discretized\n",
    "    fp_diver = FP_DivergenceExplorer(df_discretized, true_class_name = target_col, class_map={\"P\":1, \"N\":0})\n",
    "    FP_fm = fp_diver.getFrequentPatternDivergence(min_support=min_sup, metrics=[target_metric])\n",
    "    FP_fm.rename(columns = remapped_cols, inplace = True)\n",
    "    FP_fm = FP_fm[show_cols].copy()\n",
    "    FP_fm['accuracy'] = round(FP_fm['accuracy'], 5)\n",
    "    FP_fm['d_accuracy'] = round(FP_fm['d_accuracy'], 5)\n",
    "    FP_fm['t_value'] = round(FP_fm['t_value'], 2)\n",
    "    FP_fm_dict[config] = FP_fm\n",
    "    fp_divergence_dict[config] = FP_Divergence(FP_fm, target_div)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "6a460399",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of wav2vec2-base: 91.722\n",
      "Accuracy of wav2vec2-large: 93.172\n",
      "Accuracy of hubert-base: 98.418\n",
      "Accuracy of hubert-large: 98.497\n"
     ]
    }
   ],
   "source": [
    "## Check if the accuracy of the models is as expected\n",
    "for config in [\"wav2vec2-base\", \"wav2vec2-large\", \"hubert-base\", \"hubert-large\"]:\n",
    "    prediction = df_dict[config]['prediction'].sum()/len(df_dict['wav2vec2-large'])\n",
    "    print(f\"Accuracy of {config}:\", round(100*prediction,3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "f1c244c6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "wav2vec2-base\n",
      "           itemsets  accuracy  d_accuracy\n",
      "31  (gender=female)   0.92510     0.00789\n",
      "35    (gender=male)   0.90864    -0.00858\n",
      "wav2vec2-large\n",
      "           itemsets  accuracy  d_accuracy\n",
      "31  (gender=female)   0.95395     0.02223\n",
      "35    (gender=male)   0.90754    -0.02418\n",
      "hubert-base\n",
      "           itemsets  accuracy  d_accuracy\n",
      "31  (gender=female)   0.98735     0.00317\n",
      "35    (gender=male)   0.98074    -0.00344\n",
      "hubert-large\n",
      "           itemsets  accuracy  d_accuracy\n",
      "35    (gender=male)   0.98514     0.00017\n",
      "31  (gender=female)   0.98482    -0.00015\n"
     ]
    }
   ],
   "source": [
    "## Check if the accuracy of the models is as expected\n",
    "for config in [\"wav2vec2-base\", \"wav2vec2-large\", \"hubert-base\", \"hubert-large\"]:\n",
    "    print(config)\n",
    "    print(filter_itemset_df_by_attributes(FP_fm_dict[config],['gender'], alls = True)\\\n",
    "        [['itemsets', 'accuracy', 'd_accuracy']]\\\n",
    "        .sort_values('d_accuracy', ascending = False))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4341aad",
   "metadata": {},
   "source": [
    "# Divergence Wav2Vec2-Base"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "0ccb8997",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Compute the divergence for Wav2Vec2-Base\n",
    "config = 'wav2vec2-base'\n",
    "fp_divergence_i = fp_divergence_dict[config]\n",
    "\n",
    "FPdiv = fp_divergence_i.getDivergence(th_redundancy=0.0)[::-1] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "3cca7fd6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>support</th>\n",
       "      <th>itemsets</th>\n",
       "      <th>#errors</th>\n",
       "      <th>#corrects</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>d_accuracy</th>\n",
       "      <th>t_value</th>\n",
       "      <th>support_count</th>\n",
       "      <th>length</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>44217</th>\n",
       "      <td>0.03</td>\n",
       "      <td>(gender=male, speed_rate_word=high, ageRange=22-40, location=none, total_silence=high)</td>\n",
       "      <td>47</td>\n",
       "      <td>72</td>\n",
       "      <td>60.504</td>\n",
       "      <td>-31.217</td>\n",
       "      <td>7.05</td>\n",
       "      <td>119.0</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32246</th>\n",
       "      <td>0.04</td>\n",
       "      <td>(ageRange=22-40, total_silence=high, speed_rate_word=high, gender=male)</td>\n",
       "      <td>52</td>\n",
       "      <td>88</td>\n",
       "      <td>62.857</td>\n",
       "      <td>-28.864</td>\n",
       "      <td>7.13</td>\n",
       "      <td>140.0</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39743</th>\n",
       "      <td>0.03</td>\n",
       "      <td>(location=none, gender=male, trimmed_duration=low, n_words=low, total_silence=high)</td>\n",
       "      <td>45</td>\n",
       "      <td>80</td>\n",
       "      <td>64.000</td>\n",
       "      <td>-27.722</td>\n",
       "      <td>6.54</td>\n",
       "      <td>125.0</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       support  \\\n",
       "44217     0.03   \n",
       "32246     0.04   \n",
       "39743     0.03   \n",
       "\n",
       "                                                                                     itemsets  \\\n",
       "44217  (gender=male, speed_rate_word=high, ageRange=22-40, location=none, total_silence=high)   \n",
       "32246                 (ageRange=22-40, total_silence=high, speed_rate_word=high, gender=male)   \n",
       "39743     (location=none, gender=male, trimmed_duration=low, n_words=low, total_silence=high)   \n",
       "\n",
       "       #errors  #corrects  accuracy  d_accuracy  t_value  support_count  \\\n",
       "44217       47         72    60.504     -31.217     7.05          119.0   \n",
       "32246       52         88    62.857     -28.864     7.13          140.0   \n",
       "39743       45         80    64.000     -27.722     6.54          125.0   \n",
       "\n",
       "       length  \n",
       "44217       5  \n",
       "32246       4  \n",
       "39743       5  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Retrieve Most Divergent Itemsets \n",
    "n = 3\n",
    "\n",
    "from copy import deepcopy\n",
    "pr = FPdiv.head(n).copy()\n",
    "\n",
    "pr[\"support\"] = pr[\"support\"].round(2)\n",
    "pr[\"#errors\"] = pr[\"#errors\"].astype(int)\n",
    "pr[\"#corrects\"] = pr[\"#corrects\"].astype(int)\n",
    "pr[\"accuracy\"] = (pr[\"accuracy\"]*100).round(3)\n",
    "pr[\"d_accuracy\"] = (pr[\"d_accuracy\"]*100).round(3)\n",
    "\n",
    "display(pr.head(n))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "a84fc65f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>support</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>d_accuracy</th>\n",
       "      <th>t_value</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>44217</th>\n",
       "      <td>ageRange=22-40, gender=male, loc=none, speakRate=high, tot_silence=high</td>\n",
       "      <td>0.03</td>\n",
       "      <td>60.504</td>\n",
       "      <td>-31.217</td>\n",
       "      <td>7.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32246</th>\n",
       "      <td>ageRange=22-40, gender=male, speakRate=high, tot_silence=high</td>\n",
       "      <td>0.04</td>\n",
       "      <td>62.857</td>\n",
       "      <td>-28.864</td>\n",
       "      <td>7.13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39743</th>\n",
       "      <td>gender=male, loc=none, n_words=low, tot_silence=high, trim_dur=low</td>\n",
       "      <td>0.03</td>\n",
       "      <td>64.000</td>\n",
       "      <td>-27.722</td>\n",
       "      <td>6.54</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                      itemsets  \\\n",
       "44217  ageRange=22-40, gender=male, loc=none, speakRate=high, tot_silence=high   \n",
       "32246            ageRange=22-40, gender=male, speakRate=high, tot_silence=high   \n",
       "39743       gender=male, loc=none, n_words=low, tot_silence=high, trim_dur=low   \n",
       "\n",
       "       support  accuracy  d_accuracy  t_value  \n",
       "44217     0.03    60.504     -31.217     7.05  \n",
       "32246     0.04    62.857     -28.864     7.13  \n",
       "39743     0.03    64.000     -27.722     6.54  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Choose columns for better visualization \n",
    "pr_l = pr[[ \"itemsets\", \"support\", \"accuracy\", \"d_accuracy\", \"t_value\"]].copy()\n",
    "pr_l['itemsets'] = pr_l['itemsets'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "\n",
    "display(pr_l.head(3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "f8d55d60",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 97.87\n",
      "1 98.08\n",
      "2 84.44\n"
     ]
    }
   ],
   "source": [
    "for e, itemset in enumerate(pr.itemsets):\n",
    "    df_sel = slice_by_itemset(df_dict[config], itemset)\n",
    "    \n",
    "    _, percentage_problematic = percentage_problematic_speaker(df_sel)\n",
    "    print(e, round(percentage_problematic*100,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4341aad",
   "metadata": {},
   "source": [
    "# Divergence Wav2Vec2-Large"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "0ccb8997",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Compute the divergence for Wav2Vec2-Large\n",
    "config = 'wav2vec2-large'\n",
    "fp_divergence_i = fp_divergence_dict[config]\n",
    "\n",
    "FPdiv = fp_divergence_i.getDivergence(th_redundancy=0.0)[::-1] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "3cca7fd6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>support</th>\n",
       "      <th>itemsets</th>\n",
       "      <th>#errors</th>\n",
       "      <th>#corrects</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>d_accuracy</th>\n",
       "      <th>t_value</th>\n",
       "      <th>support_count</th>\n",
       "      <th>length</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>44217</th>\n",
       "      <td>0.03</td>\n",
       "      <td>(gender=male, speed_rate_word=high, ageRange=22-40, location=none, total_silence=high)</td>\n",
       "      <td>30</td>\n",
       "      <td>89</td>\n",
       "      <td>74.790</td>\n",
       "      <td>-18.382</td>\n",
       "      <td>4.72</td>\n",
       "      <td>119.0</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37151</th>\n",
       "      <td>0.03</td>\n",
       "      <td>(action=increase, speed_rate_word=high, gender=male)</td>\n",
       "      <td>33</td>\n",
       "      <td>98</td>\n",
       "      <td>74.809</td>\n",
       "      <td>-18.362</td>\n",
       "      <td>4.94</td>\n",
       "      <td>131.0</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27745</th>\n",
       "      <td>0.04</td>\n",
       "      <td>(action=activate, speed_rate_word_trimmed=low, gender=male)</td>\n",
       "      <td>38</td>\n",
       "      <td>113</td>\n",
       "      <td>74.834</td>\n",
       "      <td>-18.337</td>\n",
       "      <td>5.27</td>\n",
       "      <td>151.0</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       support  \\\n",
       "44217     0.03   \n",
       "37151     0.03   \n",
       "27745     0.04   \n",
       "\n",
       "                                                                                     itemsets  \\\n",
       "44217  (gender=male, speed_rate_word=high, ageRange=22-40, location=none, total_silence=high)   \n",
       "37151                                    (action=increase, speed_rate_word=high, gender=male)   \n",
       "27745                             (action=activate, speed_rate_word_trimmed=low, gender=male)   \n",
       "\n",
       "       #errors  #corrects  accuracy  d_accuracy  t_value  support_count  \\\n",
       "44217       30         89    74.790     -18.382     4.72          119.0   \n",
       "37151       33         98    74.809     -18.362     4.94          131.0   \n",
       "27745       38        113    74.834     -18.337     5.27          151.0   \n",
       "\n",
       "       length  \n",
       "44217       5  \n",
       "37151       3  \n",
       "27745       3  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Retrieve Most Divergent Itemsets \n",
    "n = 3\n",
    "\n",
    "from copy import deepcopy\n",
    "pr = FPdiv.head(n).copy()\n",
    "\n",
    "pr[\"support\"] = pr[\"support\"].round(2)\n",
    "pr[\"#errors\"] = pr[\"#errors\"].astype(int)\n",
    "pr[\"#corrects\"] = pr[\"#corrects\"].astype(int)\n",
    "pr[\"accuracy\"] = (pr[\"accuracy\"]*100).round(3)\n",
    "pr[\"d_accuracy\"] = (pr[\"d_accuracy\"]*100).round(3)\n",
    "\n",
    "display(pr.head(n))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "a84fc65f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>support</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>d_accuracy</th>\n",
       "      <th>t_value</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>44217</th>\n",
       "      <td>ageRange=22-40, gender=male, loc=none, speakRate=high, tot_silence=high</td>\n",
       "      <td>0.03</td>\n",
       "      <td>74.790</td>\n",
       "      <td>-18.382</td>\n",
       "      <td>4.72</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37151</th>\n",
       "      <td>action=increase, gender=male, speakRate=high</td>\n",
       "      <td>0.03</td>\n",
       "      <td>74.809</td>\n",
       "      <td>-18.362</td>\n",
       "      <td>4.94</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27745</th>\n",
       "      <td>action=activate, gender=male, speakRate_trim=low</td>\n",
       "      <td>0.04</td>\n",
       "      <td>74.834</td>\n",
       "      <td>-18.337</td>\n",
       "      <td>5.27</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                      itemsets  \\\n",
       "44217  ageRange=22-40, gender=male, loc=none, speakRate=high, tot_silence=high   \n",
       "37151                             action=increase, gender=male, speakRate=high   \n",
       "27745                         action=activate, gender=male, speakRate_trim=low   \n",
       "\n",
       "       support  accuracy  d_accuracy  t_value  \n",
       "44217     0.03    74.790     -18.382     4.72  \n",
       "37151     0.03    74.809     -18.362     4.94  \n",
       "27745     0.04    74.834     -18.337     5.27  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Choose columns for better visualization \n",
    "pr_l = pr[[ \"itemsets\", \"support\", \"accuracy\", \"d_accuracy\", \"t_value\"]].copy()\n",
    "pr_l['itemsets'] = pr_l['itemsets'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "\n",
    "display(pr_l.head(3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "f8d55d60",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 96.67\n",
      "1 54.55\n",
      "2 44.74\n"
     ]
    }
   ],
   "source": [
    "for e, itemset in enumerate(pr.itemsets):\n",
    "    df_sel = slice_by_itemset(df_dict[config], itemset)\n",
    "    \n",
    "    _, percentage_problematic = percentage_problematic_speaker(df_sel)\n",
    "    print(e, round(percentage_problematic*100,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ba25022",
   "metadata": {},
   "source": [
    "# Divergence difference Wav2Vec2-Base vs Wav2Vec2-Large"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "dc56bafe",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Compute the divergence for Wav2Vec2-Large\n",
    "config = 'wav2vec2-large'\n",
    "fp_divergence_i = fp_divergence_dict[config]\n",
    "FPdiv_w2l = fp_divergence_i.getDivergence(th_redundancy=None).copy()\n",
    "wlarge = FPdiv_w2l.set_index(\"itemsets\")\n",
    "\n",
    "## Compute the divergence for Wav2Vec2-Base\n",
    "config = 'wav2vec2-base'\n",
    "fp_divergence_i = fp_divergence_dict[config]\n",
    "FPdiv_w2b = fp_divergence_i.getDivergence(th_redundancy=None).copy()\n",
    "wbase = FPdiv_w2b.set_index(\"itemsets\")\n",
    "\n",
    "## Merge the df\n",
    "merged = wbase.join(wlarge, lsuffix='_base', rsuffix='_large')\n",
    "merged = merged.rename(columns={'support_large': 'support'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "32256e51",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Compute difference in performance between the models\n",
    "diff = \"d_difference\"\n",
    "merged[diff] = merged[\"accuracy_large\"] - merged[\"accuracy_base\"]\n",
    "merged[\"difference\"] = merged[\"accuracy_large\"] - merged[\"accuracy_base\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "2c799c4d",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Create \"Gain Base-Large W2V2\" df and compute divergence\n",
    "base_large_gain_df = merged[['support', '#errors_large', '#corrects_large', 'accuracy_large',\n",
    "       'd_accuracy_large', 't_value_large', 'support_count_large', 'length_large'] \\\n",
    "       + [diff, \"difference\", \"accuracy_base\"]]\n",
    "base_large_gain_df = base_large_gain_df.rename(columns={'length_large':'length'})\n",
    "base_large_gain_df = base_large_gain_df.reset_index()\n",
    "\n",
    "fp_divergence_difference = FP_Divergence(base_large_gain_df, diff)\n",
    "diff_nr = fp_divergence_difference.getDivergence(th_redundancy=0.0) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "5981f5a4",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Select meaningful columns \n",
    "sel = diff_nr.itemsets.values\n",
    "compare_performance = merged.loc[sel].sort_values(diff, ascending = False)\n",
    "cols = ['d_difference', 'accuracy_base', 'accuracy_large', 'support']"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c7560518",
   "metadata": {},
   "source": [
    "## Stats"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "a76357a3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Equal\n",
      "4.3573\n",
      "Greater\n",
      "63.7506\n",
      "Lower\n",
      "31.8921\n"
     ]
    }
   ],
   "source": [
    "## Compute meaningful statistics:\n",
    "diff_nr_0 = fp_divergence_difference.getDivergence(th_redundancy=None)\n",
    "\n",
    "# Percentage of itemsets for which performance are equal for the two model\n",
    "print(\"Equal\")\n",
    "print(round(100*(diff_nr_0.loc[(diff_nr_0[diff])==0].shape[0]/diff_nr_0.shape[0]),4))\n",
    "\n",
    "# Percentage of itemsets for which performance are greater for Wav2Vec2-Large\n",
    "print(\"Greater\")\n",
    "print(round(100*(diff_nr_0.loc[(diff_nr_0[diff])>0].shape[0]/diff_nr_0.shape[0]), 4))\n",
    "\n",
    "# Percentage of itemsets for which performance are lower for Wav2Vec2-Large\n",
    "print(\"Lower\")\n",
    "print(round(100*(diff_nr_0.loc[(diff_nr_0[diff])<0].shape[0]/diff_nr_0.shape[0]), 4))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "01d0c446",
   "metadata": {},
   "source": [
    "## Gain > 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "3d4c0e17",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_base</th>\n",
       "      <th>accuracy_large</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>(speed_rate_word_trimmed=low, trimmed_duration=low, location=none, action=increase, total_duration=low)</td>\n",
       "      <td>22.69</td>\n",
       "      <td>75.63</td>\n",
       "      <td>98.32</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>(trimmed_duration=low, ageRange=22-40, location=none, action=increase, total_duration=low)</td>\n",
       "      <td>22.01</td>\n",
       "      <td>70.44</td>\n",
       "      <td>92.45</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>(trimmed_duration=low, ageRange=22-40, object=volume, action=increase, total_duration=low)</td>\n",
       "      <td>21.93</td>\n",
       "      <td>70.18</td>\n",
       "      <td>92.10</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>(location=none, action=increase, total_duration=low, speed_rate_word_trimmed=low)</td>\n",
       "      <td>21.88</td>\n",
       "      <td>76.56</td>\n",
       "      <td>98.44</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>(location=none, trimmed_duration=low, ageRange=22-40, n_words=low, action=increase, total_duration=low)</td>\n",
       "      <td>20.66</td>\n",
       "      <td>69.42</td>\n",
       "      <td>90.08</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>(gender=female, location=none, trimmed_duration=low, n_words=low, action=increase)</td>\n",
       "      <td>20.32</td>\n",
       "      <td>68.29</td>\n",
       "      <td>88.62</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>(speed_rate_word_trimmed=low, action=increase, total_duration=low, trimmed_duration=low)</td>\n",
       "      <td>19.85</td>\n",
       "      <td>78.68</td>\n",
       "      <td>98.53</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>(location=none, gender=female, action=increase, total_duration=low)</td>\n",
       "      <td>19.35</td>\n",
       "      <td>74.19</td>\n",
       "      <td>93.55</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>(ageRange=22-40, object=volume, total_duration=low, trimmed_duration=low)</td>\n",
       "      <td>18.97</td>\n",
       "      <td>71.80</td>\n",
       "      <td>90.77</td>\n",
       "      <td>0.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>(ageRange=22-40, action=increase, total_duration=low, trimmed_duration=low)</td>\n",
       "      <td>18.82</td>\n",
       "      <td>74.73</td>\n",
       "      <td>93.55</td>\n",
       "      <td>0.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>(gender=male, trimmed_duration=low, ageRange=22-40, object=volume, total_duration=low)</td>\n",
       "      <td>18.66</td>\n",
       "      <td>70.90</td>\n",
       "      <td>89.55</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>(location=none, trimmed_duration=low, ageRange=22-40, n_words=low, action=increase)</td>\n",
       "      <td>18.52</td>\n",
       "      <td>69.75</td>\n",
       "      <td>88.27</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>(location=none, gender=female, action=increase, total_silence=medium)</td>\n",
       "      <td>18.42</td>\n",
       "      <td>78.07</td>\n",
       "      <td>96.49</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>(location=none, action=increase, total_duration=low, trimmed_duration=low)</td>\n",
       "      <td>18.30</td>\n",
       "      <td>72.77</td>\n",
       "      <td>91.06</td>\n",
       "      <td>0.06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>(trimmed_duration=low, ageRange=22-40, object=volume, n_words=low, total_duration=low)</td>\n",
       "      <td>18.07</td>\n",
       "      <td>71.08</td>\n",
       "      <td>89.16</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>(trimmed_duration=low, ageRange=22-40, object=volume, n_words=low, action=increase)</td>\n",
       "      <td>17.97</td>\n",
       "      <td>69.53</td>\n",
       "      <td>87.50</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>(ageRange=22-40, object=volume, trimmed_duration=low, action=increase)</td>\n",
       "      <td>17.90</td>\n",
       "      <td>71.60</td>\n",
       "      <td>89.51</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>(gender=female, speed_rate_word=high, trimmed_duration=low, location=none, action=increase)</td>\n",
       "      <td>17.80</td>\n",
       "      <td>70.34</td>\n",
       "      <td>88.14</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>(speed_rate_word=high, trimmed_duration=low, ageRange=22-40, object=volume, n_words=low)</td>\n",
       "      <td>17.75</td>\n",
       "      <td>66.86</td>\n",
       "      <td>84.62</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>(gender=male, trimmed_duration=low, ageRange=22-40, location=none, action=increase)</td>\n",
       "      <td>17.69</td>\n",
       "      <td>69.23</td>\n",
       "      <td>86.92</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                   itemsets  \\\n",
       "0   (speed_rate_word_trimmed=low, trimmed_duration=low, location=none, action=increase, total_duration=low)   \n",
       "1                (trimmed_duration=low, ageRange=22-40, location=none, action=increase, total_duration=low)   \n",
       "2                (trimmed_duration=low, ageRange=22-40, object=volume, action=increase, total_duration=low)   \n",
       "3                         (location=none, action=increase, total_duration=low, speed_rate_word_trimmed=low)   \n",
       "4   (location=none, trimmed_duration=low, ageRange=22-40, n_words=low, action=increase, total_duration=low)   \n",
       "5                        (gender=female, location=none, trimmed_duration=low, n_words=low, action=increase)   \n",
       "6                  (speed_rate_word_trimmed=low, action=increase, total_duration=low, trimmed_duration=low)   \n",
       "7                                       (location=none, gender=female, action=increase, total_duration=low)   \n",
       "8                                 (ageRange=22-40, object=volume, total_duration=low, trimmed_duration=low)   \n",
       "9                               (ageRange=22-40, action=increase, total_duration=low, trimmed_duration=low)   \n",
       "10                   (gender=male, trimmed_duration=low, ageRange=22-40, object=volume, total_duration=low)   \n",
       "11                      (location=none, trimmed_duration=low, ageRange=22-40, n_words=low, action=increase)   \n",
       "12                                    (location=none, gender=female, action=increase, total_silence=medium)   \n",
       "13                               (location=none, action=increase, total_duration=low, trimmed_duration=low)   \n",
       "14                   (trimmed_duration=low, ageRange=22-40, object=volume, n_words=low, total_duration=low)   \n",
       "15                      (trimmed_duration=low, ageRange=22-40, object=volume, n_words=low, action=increase)   \n",
       "16                                   (ageRange=22-40, object=volume, trimmed_duration=low, action=increase)   \n",
       "17              (gender=female, speed_rate_word=high, trimmed_duration=low, location=none, action=increase)   \n",
       "18                 (speed_rate_word=high, trimmed_duration=low, ageRange=22-40, object=volume, n_words=low)   \n",
       "19                      (gender=male, trimmed_duration=low, ageRange=22-40, location=none, action=increase)   \n",
       "\n",
       "    d_difference  accuracy_base  accuracy_large  support  \n",
       "0          22.69          75.63           98.32     0.03  \n",
       "1          22.01          70.44           92.45     0.04  \n",
       "2          21.93          70.18           92.10     0.03  \n",
       "3          21.88          76.56           98.44     0.03  \n",
       "4          20.66          69.42           90.08     0.03  \n",
       "5          20.32          68.29           88.62     0.03  \n",
       "6          19.85          78.68           98.53     0.04  \n",
       "7          19.35          74.19           93.55     0.03  \n",
       "8          18.97          71.80           90.77     0.05  \n",
       "9          18.82          74.73           93.55     0.05  \n",
       "10         18.66          70.90           89.55     0.04  \n",
       "11         18.52          69.75           88.27     0.04  \n",
       "12         18.42          78.07           96.49     0.03  \n",
       "13         18.30          72.77           91.06     0.06  \n",
       "14         18.07          71.08           89.16     0.04  \n",
       "15         17.97          69.53           87.50     0.03  \n",
       "16         17.90          71.60           89.51     0.04  \n",
       "17         17.80          70.34           88.14     0.03  \n",
       "18         17.75          66.86           84.62     0.04  \n",
       "19         17.69          69.23           86.92     0.03  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Retrieve the data cohorts for which Wav2Vec2-Large performs better than Wav2Vec2-Base\n",
    "pr = compare_performance[cols].head(20).reset_index()\n",
    "pr[\"support\"] = pr[\"support\"].round(2)\n",
    "pr[\"accuracy_large\"] = (pr[\"accuracy_large\"]*100).round(2)\n",
    "pr[\"accuracy_base\"] = (pr[\"accuracy_base\"]*100).round(2)\n",
    "pr[\"d_difference\"] = (pr[\"d_difference\"]*100).round(2)\n",
    "\n",
    "display(pr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "86746d6b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_base</th>\n",
       "      <th>accuracy_large</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>action=increase, loc=none, speakRate_trim=low, dur=low, trim_dur=low</td>\n",
       "      <td>22.69</td>\n",
       "      <td>75.63</td>\n",
       "      <td>98.32</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>action=increase, ageRange=22-40, loc=none, dur=low, trim_dur=low</td>\n",
       "      <td>22.01</td>\n",
       "      <td>70.44</td>\n",
       "      <td>92.45</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                               itemsets  \\\n",
       "0  action=increase, loc=none, speakRate_trim=low, dur=low, trim_dur=low   \n",
       "1      action=increase, ageRange=22-40, loc=none, dur=low, trim_dur=low   \n",
       "\n",
       "   d_difference  accuracy_base  accuracy_large  support  \n",
       "0         22.69          75.63           98.32     0.03  \n",
       "1         22.01          70.44           92.45     0.04  "
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Abbreviate itemset names for better visualization\n",
    "pr_l = pr.head(2).copy()\n",
    "pr_l['itemsets'] = pr_l['itemsets'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "pr_l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "2d4d940c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 200x200 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Compute Shapley Values for a given itemset\n",
    "itemset_1 = pr.iloc[1].itemsets\n",
    "itemset_shap = fp_divergence_difference.computeShapleyValue(itemset_1)\n",
    "fp_divergence_difference.plotShapleyValue(shapley_values=abbreviateDict(itemset_shap, abbreviations), \n",
    "                                          sizeFig=(2,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b4b34610",
   "metadata": {},
   "source": [
    "## Gain < 0 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "4ac034d4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_base</th>\n",
       "      <th>accuracy_large</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>(action=activate, speed_rate_word=low, gender=male)</td>\n",
       "      <td>-20.97</td>\n",
       "      <td>96.77</td>\n",
       "      <td>75.81</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>(action=activate, speed_rate_word_trimmed=low, gender=male)</td>\n",
       "      <td>-16.56</td>\n",
       "      <td>91.39</td>\n",
       "      <td>74.83</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>(action=activate, n_words=medium, gender=male)</td>\n",
       "      <td>-16.39</td>\n",
       "      <td>94.54</td>\n",
       "      <td>78.14</td>\n",
       "      <td>0.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>(location=washroom, action=activate, n_words=medium)</td>\n",
       "      <td>-16.26</td>\n",
       "      <td>96.75</td>\n",
       "      <td>80.49</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>(speed_rate_word=low, object=lights, action=activate, n_words=medium)</td>\n",
       "      <td>-16.13</td>\n",
       "      <td>99.19</td>\n",
       "      <td>83.06</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>(action=activate, trimmed_duration=medium, gender=male)</td>\n",
       "      <td>-16.03</td>\n",
       "      <td>95.42</td>\n",
       "      <td>79.39</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>(object=lights, location=washroom, gender=male)</td>\n",
       "      <td>-15.65</td>\n",
       "      <td>95.65</td>\n",
       "      <td>80.00</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>(gender=male, action=activate, total_silence=low)</td>\n",
       "      <td>-15.54</td>\n",
       "      <td>93.92</td>\n",
       "      <td>78.38</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>(object=lights, action=activate, trimmed_duration=medium, gender=male)</td>\n",
       "      <td>-15.45</td>\n",
       "      <td>95.12</td>\n",
       "      <td>79.68</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>(object=lights, action=activate, speed_rate_word=low, total_silence=low)</td>\n",
       "      <td>-15.44</td>\n",
       "      <td>96.32</td>\n",
       "      <td>80.88</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>(ageRange=22-40, action=activate, n_words=medium, gender=male)</td>\n",
       "      <td>-15.44</td>\n",
       "      <td>93.38</td>\n",
       "      <td>77.94</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>(speed_rate_word=low, action=activate, n_words=medium)</td>\n",
       "      <td>-15.17</td>\n",
       "      <td>97.93</td>\n",
       "      <td>82.76</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>(object=lights, action=activate, n_words=medium, gender=male)</td>\n",
       "      <td>-15.13</td>\n",
       "      <td>94.08</td>\n",
       "      <td>78.95</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>(gender=male, ageRange=22-40, action=activate, total_silence=low)</td>\n",
       "      <td>-14.79</td>\n",
       "      <td>94.37</td>\n",
       "      <td>79.58</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>(object=lights, gender=male, speed_rate_word=low, total_silence=low)</td>\n",
       "      <td>-14.69</td>\n",
       "      <td>97.90</td>\n",
       "      <td>83.22</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>(ageRange=22-40, action=activate, speed_rate_word=low)</td>\n",
       "      <td>-14.40</td>\n",
       "      <td>95.20</td>\n",
       "      <td>80.80</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>(location=washroom, action=activate)</td>\n",
       "      <td>-14.19</td>\n",
       "      <td>96.77</td>\n",
       "      <td>82.58</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>(speed_rate_word=low, action=activate, n_words=medium, total_silence=low)</td>\n",
       "      <td>-14.06</td>\n",
       "      <td>97.66</td>\n",
       "      <td>83.59</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>(object=lights, action=activate, n_words=medium, total_silence=low)</td>\n",
       "      <td>-14.05</td>\n",
       "      <td>98.35</td>\n",
       "      <td>84.30</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>(action=activate, n_words=medium, total_silence=low)</td>\n",
       "      <td>-13.99</td>\n",
       "      <td>97.20</td>\n",
       "      <td>83.22</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                     itemsets  \\\n",
       "0                         (action=activate, speed_rate_word=low, gender=male)   \n",
       "1                 (action=activate, speed_rate_word_trimmed=low, gender=male)   \n",
       "2                              (action=activate, n_words=medium, gender=male)   \n",
       "3                        (location=washroom, action=activate, n_words=medium)   \n",
       "4       (speed_rate_word=low, object=lights, action=activate, n_words=medium)   \n",
       "5                     (action=activate, trimmed_duration=medium, gender=male)   \n",
       "6                             (object=lights, location=washroom, gender=male)   \n",
       "7                           (gender=male, action=activate, total_silence=low)   \n",
       "8      (object=lights, action=activate, trimmed_duration=medium, gender=male)   \n",
       "9    (object=lights, action=activate, speed_rate_word=low, total_silence=low)   \n",
       "10             (ageRange=22-40, action=activate, n_words=medium, gender=male)   \n",
       "11                     (speed_rate_word=low, action=activate, n_words=medium)   \n",
       "12              (object=lights, action=activate, n_words=medium, gender=male)   \n",
       "13          (gender=male, ageRange=22-40, action=activate, total_silence=low)   \n",
       "14       (object=lights, gender=male, speed_rate_word=low, total_silence=low)   \n",
       "15                     (ageRange=22-40, action=activate, speed_rate_word=low)   \n",
       "16                                       (location=washroom, action=activate)   \n",
       "17  (speed_rate_word=low, action=activate, n_words=medium, total_silence=low)   \n",
       "18        (object=lights, action=activate, n_words=medium, total_silence=low)   \n",
       "19                       (action=activate, n_words=medium, total_silence=low)   \n",
       "\n",
       "    d_difference  accuracy_base  accuracy_large  support  \n",
       "0         -20.97          96.77           75.81     0.03  \n",
       "1         -16.56          91.39           74.83     0.04  \n",
       "2         -16.39          94.54           78.14     0.05  \n",
       "3         -16.26          96.75           80.49     0.03  \n",
       "4         -16.13          99.19           83.06     0.03  \n",
       "5         -16.03          95.42           79.39     0.03  \n",
       "6         -15.65          95.65           80.00     0.03  \n",
       "7         -15.54          93.92           78.38     0.04  \n",
       "8         -15.45          95.12           79.68     0.03  \n",
       "9         -15.44          96.32           80.88     0.04  \n",
       "10        -15.44          93.38           77.94     0.04  \n",
       "11        -15.17          97.93           82.76     0.04  \n",
       "12        -15.13          94.08           78.95     0.04  \n",
       "13        -14.79          94.37           79.58     0.04  \n",
       "14        -14.69          97.90           83.22     0.04  \n",
       "15        -14.40          95.20           80.80     0.03  \n",
       "16        -14.19          96.77           82.58     0.04  \n",
       "17        -14.06          97.66           83.59     0.03  \n",
       "18        -14.05          98.35           84.30     0.03  \n",
       "19        -13.99          97.20           83.22     0.04  "
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Retrieve the data cohorts for which Wav2Vec2-Large performs worse than Wav2Vec2-Base\n",
    "pr = compare_performance[cols][::-1].head(20).reset_index()\n",
    "pr[\"support\"] = pr[\"support\"].round(2)\n",
    "pr[\"accuracy_large\"] = (pr[\"accuracy_large\"]*100).round(2)\n",
    "pr[\"accuracy_base\"] = (pr[\"accuracy_base\"]*100).round(2)\n",
    "pr[\"d_difference\"] = (pr[\"d_difference\"]*100).round(2)\n",
    "\n",
    "pr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "7bd297bd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_base</th>\n",
       "      <th>accuracy_large</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>action=activate, gender=male, speakRate=low</td>\n",
       "      <td>-20.97</td>\n",
       "      <td>96.77</td>\n",
       "      <td>75.81</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>action=activate, gender=male, speakRate_trim=low</td>\n",
       "      <td>-16.56</td>\n",
       "      <td>91.39</td>\n",
       "      <td>74.83</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                           itemsets  d_difference  \\\n",
       "0       action=activate, gender=male, speakRate=low        -20.97   \n",
       "1  action=activate, gender=male, speakRate_trim=low        -16.56   \n",
       "\n",
       "   accuracy_base  accuracy_large  support  \n",
       "0          96.77           75.81     0.03  \n",
       "1          91.39           74.83     0.04  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Abbreviate itemset names for better visualization\n",
    "pr_l = pr.head(2).copy()\n",
    "pr_l['itemsets'] = pr_l['itemsets'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "pr_l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "96b57661",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 200x200 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Compute Shapley Values for a given itemset\n",
    "itemset_1 = pr.iloc[0].itemsets\n",
    "itemset_shap = fp_divergence_difference.computeShapleyValue(itemset_1)\n",
    "fp_divergence_difference.plotShapleyValue(shapley_values=abbreviateDict(itemset_shap, abbreviations), \n",
    "                                          sizeFig=(2,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a566eb0",
   "metadata": {},
   "source": [
    "## Gain = 0 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "d6fb950a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_base</th>\n",
       "      <th>accuracy_large</th>\n",
       "      <th>support</th>\n",
       "      <th>support_large</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3167</th>\n",
       "      <td>(n_words=low, action=increase, speed_rate_word=high, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>75.41</td>\n",
       "      <td>75.41</td>\n",
       "      <td>0.032165</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3166</th>\n",
       "      <td>(n_words=low, object=volume, action=increase, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>78.15</td>\n",
       "      <td>78.15</td>\n",
       "      <td>0.031374</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3165</th>\n",
       "      <td>(ageRange=22-40, action=activate, total_silence=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>84.77</td>\n",
       "      <td>84.77</td>\n",
       "      <td>0.039810</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3164</th>\n",
       "      <td>(n_words=low, action=activate, total_duration=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>84.85</td>\n",
       "      <td>84.85</td>\n",
       "      <td>0.034801</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3163</th>\n",
       "      <td>(total_duration=low, total_silence=medium, speed_rate_word=high, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>86.86</td>\n",
       "      <td>86.86</td>\n",
       "      <td>0.036119</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3162</th>\n",
       "      <td>(location=none, total_silence=medium, speed_rate_word=high, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>87.16</td>\n",
       "      <td>87.16</td>\n",
       "      <td>0.039019</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3161</th>\n",
       "      <td>(n_words=low, total_duration=low, total_silence=medium, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>87.34</td>\n",
       "      <td>87.34</td>\n",
       "      <td>0.041656</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3160</th>\n",
       "      <td>(action=increase, trimmed_duration=medium, n_words=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>87.72</td>\n",
       "      <td>87.72</td>\n",
       "      <td>0.030055</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3159</th>\n",
       "      <td>(speed_rate_word_trimmed=medium, total_duration=low, total_silence=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>88.19</td>\n",
       "      <td>88.19</td>\n",
       "      <td>0.033483</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3158</th>\n",
       "      <td>(location=none, speed_rate_word_trimmed=medium, total_duration=low, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>88.33</td>\n",
       "      <td>88.33</td>\n",
       "      <td>0.047456</td>\n",
       "      <td>0.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3157</th>\n",
       "      <td>(n_words=low, action=activate, speed_rate_word_trimmed=high, speed_rate_word=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>88.98</td>\n",
       "      <td>88.98</td>\n",
       "      <td>0.033483</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3156</th>\n",
       "      <td>(action=increase, total_silence=medium, ageRange=41-65)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>89.17</td>\n",
       "      <td>89.17</td>\n",
       "      <td>0.031637</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3155</th>\n",
       "      <td>(ageRange=22-40, action=increase, trimmed_duration=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>90.00</td>\n",
       "      <td>90.00</td>\n",
       "      <td>0.036910</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3154</th>\n",
       "      <td>(speed_rate_word_trimmed=low, speed_rate_word=medium, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>90.05</td>\n",
       "      <td>90.05</td>\n",
       "      <td>0.055629</td>\n",
       "      <td>0.06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3153</th>\n",
       "      <td>(ageRange=22-40, gender=female, total_silence=high, trimmed_duration=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>91.13</td>\n",
       "      <td>91.13</td>\n",
       "      <td>0.032692</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3152</th>\n",
       "      <td>(n_words=low, gender=female, action=decrease)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>91.37</td>\n",
       "      <td>91.37</td>\n",
       "      <td>0.051938</td>\n",
       "      <td>0.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3151</th>\n",
       "      <td>(ageRange=22-40, trimmed_duration=medium, total_silence=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>91.40</td>\n",
       "      <td>91.40</td>\n",
       "      <td>0.058265</td>\n",
       "      <td>0.06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3150</th>\n",
       "      <td>(total_duration=medium, speed_rate_word_trimmed=high, speed_rate_word=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>91.44</td>\n",
       "      <td>91.44</td>\n",
       "      <td>0.067756</td>\n",
       "      <td>0.07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3149</th>\n",
       "      <td>(n_words=low, total_duration=medium, speed_rate_word_trimmed=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>91.44</td>\n",
       "      <td>91.44</td>\n",
       "      <td>0.067756</td>\n",
       "      <td>0.07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3148</th>\n",
       "      <td>(speed_rate_word_trimmed=medium, total_duration=low, trimmed_duration=medium, total_silence=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>91.45</td>\n",
       "      <td>91.45</td>\n",
       "      <td>0.030846</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                              itemsets  \\\n",
       "3167                                 (n_words=low, action=increase, speed_rate_word=high, gender=male)   \n",
       "3166                                        (n_words=low, object=volume, action=increase, gender=male)   \n",
       "3165                                             (ageRange=22-40, action=activate, total_silence=high)   \n",
       "3164                                                (n_words=low, action=activate, total_duration=low)   \n",
       "3163                     (total_duration=low, total_silence=medium, speed_rate_word=high, gender=male)   \n",
       "3162                          (location=none, total_silence=medium, speed_rate_word=high, gender=male)   \n",
       "3161                              (n_words=low, total_duration=low, total_silence=medium, gender=male)   \n",
       "3160                                        (action=increase, trimmed_duration=medium, n_words=medium)   \n",
       "3159                        (speed_rate_word_trimmed=medium, total_duration=low, total_silence=medium)   \n",
       "3158                  (location=none, speed_rate_word_trimmed=medium, total_duration=low, gender=male)   \n",
       "3157                (n_words=low, action=activate, speed_rate_word_trimmed=high, speed_rate_word=high)   \n",
       "3156                                           (action=increase, total_silence=medium, ageRange=41-65)   \n",
       "3155                                        (ageRange=22-40, action=increase, trimmed_duration=medium)   \n",
       "3154                                (speed_rate_word_trimmed=low, speed_rate_word=medium, gender=male)   \n",
       "3153                         (ageRange=22-40, gender=female, total_silence=high, trimmed_duration=low)   \n",
       "3152                                                     (n_words=low, gender=female, action=decrease)   \n",
       "3151                                     (ageRange=22-40, trimmed_duration=medium, total_silence=high)   \n",
       "3150                       (total_duration=medium, speed_rate_word_trimmed=high, speed_rate_word=high)   \n",
       "3149                                (n_words=low, total_duration=medium, speed_rate_word_trimmed=high)   \n",
       "3148  (speed_rate_word_trimmed=medium, total_duration=low, trimmed_duration=medium, total_silence=low)   \n",
       "\n",
       "      d_difference  accuracy_base  accuracy_large   support  support_large  \n",
       "3167           0.0          75.41           75.41  0.032165           0.03  \n",
       "3166           0.0          78.15           78.15  0.031374           0.03  \n",
       "3165           0.0          84.77           84.77  0.039810           0.04  \n",
       "3164           0.0          84.85           84.85  0.034801           0.03  \n",
       "3163           0.0          86.86           86.86  0.036119           0.04  \n",
       "3162           0.0          87.16           87.16  0.039019           0.04  \n",
       "3161           0.0          87.34           87.34  0.041656           0.04  \n",
       "3160           0.0          87.72           87.72  0.030055           0.03  \n",
       "3159           0.0          88.19           88.19  0.033483           0.03  \n",
       "3158           0.0          88.33           88.33  0.047456           0.05  \n",
       "3157           0.0          88.98           88.98  0.033483           0.03  \n",
       "3156           0.0          89.17           89.17  0.031637           0.03  \n",
       "3155           0.0          90.00           90.00  0.036910           0.04  \n",
       "3154           0.0          90.05           90.05  0.055629           0.06  \n",
       "3153           0.0          91.13           91.13  0.032692           0.03  \n",
       "3152           0.0          91.37           91.37  0.051938           0.05  \n",
       "3151           0.0          91.40           91.40  0.058265           0.06  \n",
       "3150           0.0          91.44           91.44  0.067756           0.07  \n",
       "3149           0.0          91.44           91.44  0.067756           0.07  \n",
       "3148           0.0          91.45           91.45  0.030846           0.03  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Retrieve the data cohorts for which Wav2Vec2-Large performs equal to Wav2Vec2-Base\n",
    "pr = merged.loc[ fp_divergence_difference.getDivergence(th_redundancy=0.0).itemsets.values][cols].reset_index()\n",
    "pr[\"support_large\"] = pr[\"support\"].round(2)\n",
    "pr[\"accuracy_large\"] = (pr[\"accuracy_large\"]*100).round(2)\n",
    "pr[\"accuracy_base\"] = (pr[\"accuracy_base\"]*100).round(2)\n",
    "pr[\"d_difference\"] = (pr[\"d_difference\"]*100).round(2)\n",
    "pr = pr.loc[abs(pr[\"d_difference\"])==0]\n",
    "\n",
    "pr = pr.sort_values(\"d_difference\").sort_values(\"accuracy_large\")\n",
    "pr.head(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "77fef497",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_base</th>\n",
       "      <th>accuracy_large</th>\n",
       "      <th>support</th>\n",
       "      <th>support_large</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3167</th>\n",
       "      <td>action=increase, gender=male, n_words=low, speakRate=high</td>\n",
       "      <td>0.0</td>\n",
       "      <td>75.41</td>\n",
       "      <td>75.41</td>\n",
       "      <td>0.032165</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3166</th>\n",
       "      <td>action=increase, gender=male, n_words=low, object=volume</td>\n",
       "      <td>0.0</td>\n",
       "      <td>78.15</td>\n",
       "      <td>78.15</td>\n",
       "      <td>0.031374</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       itemsets  d_difference  \\\n",
       "3167  action=increase, gender=male, n_words=low, speakRate=high           0.0   \n",
       "3166   action=increase, gender=male, n_words=low, object=volume           0.0   \n",
       "\n",
       "      accuracy_base  accuracy_large   support  support_large  \n",
       "3167          75.41           75.41  0.032165           0.03  \n",
       "3166          78.15           78.15  0.031374           0.03  "
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Abbreviate itemset names for better visualization\n",
    "pr_l = pr.head(2).copy()\n",
    "pr_l['itemsets'] = pr_l['itemsets'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "pr_l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "d17cfdc7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 200x200 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Compute Shapley Values for a given itemset\n",
    "itemset_1 = pr.iloc[0].itemsets\n",
    "itemset_shap = fp_divergence_difference.computeShapleyValue(itemset_1)\n",
    "fp_divergence_difference.plotShapleyValue(shapley_values=abbreviateDict(itemset_shap, abbreviations), \n",
    "                                          sizeFig=(2,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9707900a",
   "metadata": {},
   "source": [
    "## Inspect gender"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b969a5e2",
   "metadata": {},
   "source": [
    "We consider subgroups among the top-10 of each gain category for which both two genders are above support."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "3464ced3",
   "metadata": {},
   "outputs": [],
   "source": [
    "n = 20\n",
    "cols = ['d_difference','accuracy_base', 'accuracy_large', 'support']\n",
    "comparison_res = []\n",
    "config1 = \"wav2vec2-base\"\n",
    "config2 = \"wav2vec2-large\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "56047630",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Gain</th>\n",
       "      <th>Subgroup</th>\n",
       "      <th>wav2vec2-base_gender=female</th>\n",
       "      <th>wav2vec2-large_gender=female</th>\n",
       "      <th>sup_gender=female</th>\n",
       "      <th>wav2vec2-base_gender=male</th>\n",
       "      <th>wav2vec2-large_gender=male</th>\n",
       "      <th>sup_gender=male</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(n_words=low, location=none, action=increase, trimmed_duration=low)</td>\n",
       "      <td>68.29</td>\n",
       "      <td>88.62</td>\n",
       "      <td>123.0</td>\n",
       "      <td>73.48</td>\n",
       "      <td>81.82</td>\n",
       "      <td>132.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(location=none, action=increase, total_duration=low)</td>\n",
       "      <td>74.19</td>\n",
       "      <td>93.55</td>\n",
       "      <td>124.0</td>\n",
       "      <td>78.74</td>\n",
       "      <td>89.08</td>\n",
       "      <td>174.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>=</td>\n",
       "      <td>(n_words=low, action=increase, speed_rate_word=high)</td>\n",
       "      <td>78.41</td>\n",
       "      <td>90.91</td>\n",
       "      <td>176.0</td>\n",
       "      <td>75.41</td>\n",
       "      <td>75.41</td>\n",
       "      <td>122.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>=</td>\n",
       "      <td>(n_words=low, object=volume, action=increase)</td>\n",
       "      <td>73.44</td>\n",
       "      <td>89.84</td>\n",
       "      <td>128.0</td>\n",
       "      <td>78.15</td>\n",
       "      <td>78.15</td>\n",
       "      <td>119.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>=</td>\n",
       "      <td>(total_duration=low, total_silence=medium, speed_rate_word=high)</td>\n",
       "      <td>85.71</td>\n",
       "      <td>95.65</td>\n",
       "      <td>161.0</td>\n",
       "      <td>86.86</td>\n",
       "      <td>86.86</td>\n",
       "      <td>137.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>=</td>\n",
       "      <td>(location=none, total_silence=medium, speed_rate_word=high)</td>\n",
       "      <td>90.08</td>\n",
       "      <td>95.80</td>\n",
       "      <td>262.0</td>\n",
       "      <td>87.16</td>\n",
       "      <td>87.16</td>\n",
       "      <td>148.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>=</td>\n",
       "      <td>(n_words=low, total_duration=low, total_silence=medium)</td>\n",
       "      <td>85.55</td>\n",
       "      <td>95.38</td>\n",
       "      <td>173.0</td>\n",
       "      <td>87.34</td>\n",
       "      <td>87.34</td>\n",
       "      <td>158.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>=</td>\n",
       "      <td>(location=none, speed_rate_word_trimmed=medium, total_duration=low)</td>\n",
       "      <td>84.13</td>\n",
       "      <td>94.44</td>\n",
       "      <td>126.0</td>\n",
       "      <td>88.33</td>\n",
       "      <td>88.33</td>\n",
       "      <td>180.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>=</td>\n",
       "      <td>(speed_rate_word_trimmed=low, speed_rate_word=medium)</td>\n",
       "      <td>91.82</td>\n",
       "      <td>94.34</td>\n",
       "      <td>159.0</td>\n",
       "      <td>90.05</td>\n",
       "      <td>90.05</td>\n",
       "      <td>211.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>=</td>\n",
       "      <td>(n_words=low, action=decrease)</td>\n",
       "      <td>91.37</td>\n",
       "      <td>91.37</td>\n",
       "      <td>197.0</td>\n",
       "      <td>82.90</td>\n",
       "      <td>92.75</td>\n",
       "      <td>193.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(action=activate, n_words=medium)</td>\n",
       "      <td>96.17</td>\n",
       "      <td>95.22</td>\n",
       "      <td>209.0</td>\n",
       "      <td>94.54</td>\n",
       "      <td>78.14</td>\n",
       "      <td>183.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(action=activate, trimmed_duration=medium)</td>\n",
       "      <td>97.71</td>\n",
       "      <td>94.66</td>\n",
       "      <td>131.0</td>\n",
       "      <td>95.42</td>\n",
       "      <td>79.39</td>\n",
       "      <td>131.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(object=lights, location=washroom)</td>\n",
       "      <td>96.58</td>\n",
       "      <td>96.58</td>\n",
       "      <td>117.0</td>\n",
       "      <td>95.65</td>\n",
       "      <td>80.00</td>\n",
       "      <td>115.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(object=lights, action=activate, n_words=medium)</td>\n",
       "      <td>97.63</td>\n",
       "      <td>95.27</td>\n",
       "      <td>169.0</td>\n",
       "      <td>94.08</td>\n",
       "      <td>78.95</td>\n",
       "      <td>152.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Gain                                                             Subgroup  \\\n",
       "0    >  (n_words=low, location=none, action=increase, trimmed_duration=low)   \n",
       "1    >                 (location=none, action=increase, total_duration=low)   \n",
       "0    =                 (n_words=low, action=increase, speed_rate_word=high)   \n",
       "1    =                        (n_words=low, object=volume, action=increase)   \n",
       "2    =     (total_duration=low, total_silence=medium, speed_rate_word=high)   \n",
       "3    =          (location=none, total_silence=medium, speed_rate_word=high)   \n",
       "4    =              (n_words=low, total_duration=low, total_silence=medium)   \n",
       "5    =  (location=none, speed_rate_word_trimmed=medium, total_duration=low)   \n",
       "6    =                (speed_rate_word_trimmed=low, speed_rate_word=medium)   \n",
       "7    =                                       (n_words=low, action=decrease)   \n",
       "0    <                                    (action=activate, n_words=medium)   \n",
       "1    <                           (action=activate, trimmed_duration=medium)   \n",
       "2    <                                   (object=lights, location=washroom)   \n",
       "3    <                     (object=lights, action=activate, n_words=medium)   \n",
       "\n",
       "   wav2vec2-base_gender=female  wav2vec2-large_gender=female  \\\n",
       "0                        68.29                         88.62   \n",
       "1                        74.19                         93.55   \n",
       "0                        78.41                         90.91   \n",
       "1                        73.44                         89.84   \n",
       "2                        85.71                         95.65   \n",
       "3                        90.08                         95.80   \n",
       "4                        85.55                         95.38   \n",
       "5                        84.13                         94.44   \n",
       "6                        91.82                         94.34   \n",
       "7                        91.37                         91.37   \n",
       "0                        96.17                         95.22   \n",
       "1                        97.71                         94.66   \n",
       "2                        96.58                         96.58   \n",
       "3                        97.63                         95.27   \n",
       "\n",
       "   sup_gender=female  wav2vec2-base_gender=male  wav2vec2-large_gender=male  \\\n",
       "0              123.0                      73.48                       81.82   \n",
       "1              124.0                      78.74                       89.08   \n",
       "0              176.0                      75.41                       75.41   \n",
       "1              128.0                      78.15                       78.15   \n",
       "2              161.0                      86.86                       86.86   \n",
       "3              262.0                      87.16                       87.16   \n",
       "4              173.0                      87.34                       87.34   \n",
       "5              126.0                      88.33                       88.33   \n",
       "6              159.0                      90.05                       90.05   \n",
       "7              197.0                      82.90                       92.75   \n",
       "0              209.0                      94.54                       78.14   \n",
       "1              131.0                      95.42                       79.39   \n",
       "2              117.0                      95.65                       80.00   \n",
       "3              169.0                      94.08                       78.95   \n",
       "\n",
       "   sup_gender=male  \n",
       "0            132.0  \n",
       "1            174.0  \n",
       "0            122.0  \n",
       "1            119.0  \n",
       "2            137.0  \n",
       "3            148.0  \n",
       "4            158.0  \n",
       "5            180.0  \n",
       "6            211.0  \n",
       "7            193.0  \n",
       "0            183.0  \n",
       "1            131.0  \n",
       "2            115.0  \n",
       "3            152.0  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Compute gain in performance when size changes, when considering different genders\n",
    "great = compare_performance[cols].head(n).reset_index()\n",
    "notr = merged.loc[ fp_divergence_difference.getDivergence(th_redundancy=0.0).itemsets.values][cols].reset_index()\n",
    "equal = notr.loc[notr[diff]==0].sort_values('accuracy_large').head(n)\n",
    "low = compare_performance[cols][::-1].head(n).reset_index()\n",
    "\n",
    "for gain_type, pr in zip(['>', \"=\", \"<\"], [great, equal, low]):\n",
    "    gain_df_gender = filter_itemset_df_by_attributes(pr, ['gender'], alls=False)\n",
    "\n",
    "    itemsets = gain_df_gender.itemsets.values\n",
    "\n",
    "    config1 = \"wav2vec2-base\"\n",
    "    config2 = \"wav2vec2-large\"\n",
    "    get_accuracy = lambda dfi: sum(dfi['prediction'].values.astype(int))/len(dfi)\n",
    "    get_support_count = lambda dfi: len(dfi)\n",
    "\n",
    "    res2 = []\n",
    "    res = []\n",
    "\n",
    "    genders = ['gender=female', 'gender=male']\n",
    "    configs = [config1, config2]\n",
    "    for itemset_i in itemsets:\n",
    "        itemset_i = itemset_i.difference(genders)\n",
    "        r2 = [gain_type, itemset_i]\n",
    "        both_genders = []\n",
    "        both_genders_2 = []\n",
    "        for gender in genders:\n",
    "            itemset = itemset_i.union([gender])\n",
    "            r = [gain_type, itemset]\n",
    "\n",
    "            for config in configs:\n",
    "                if itemset in  fp_divergence_dict[config].itemset_divergence[len(itemset)]:\n",
    "                    info = fp_divergence_dict[config].getInfoItemset(itemset).iloc[0]\n",
    "                    acc = info[\"accuracy\"]\n",
    "                    s = info[\"support_count\"]\n",
    "\n",
    "                    r.append((acc*100).round(2))\n",
    "                    r2.append((acc*100).round(2))\n",
    "                    r.append(s)\n",
    "\n",
    "            if len(r)>2:\n",
    "                r2.append(s)\n",
    "                both_genders.append(r)\n",
    "                both_genders_2.append(r2)\n",
    "\n",
    "        if len(both_genders)==2:\n",
    "            for r in both_genders:\n",
    "                res.append(r)\n",
    "            res2.append(r2)\n",
    "    \n",
    "    cols =[]\n",
    "    for gender in genders:\n",
    "        cols.extend([f\"{c}_{gender}\" for c in configs + [\"sup\"]])\n",
    "    v = pd.DataFrame(res, columns = ['Gain',\"Subgroup\"] + configs + [\"S1\", \"S2\"])\n",
    "    v2 = pd.DataFrame(res2, columns = ['Gain', \"Subgroup\"] + cols )\n",
    "\n",
    "    comparison_res.append(v2)\n",
    "\n",
    "comparison_res = pd.concat(comparison_res)\n",
    "comparison_res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "c632adb1",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Gain</th>\n",
       "      <th>Subgroup_str</th>\n",
       "      <th>wav2vec2-base_gender=female</th>\n",
       "      <th>wav2vec2-large_gender=female</th>\n",
       "      <th>wav2vec2-base_gender=male</th>\n",
       "      <th>wav2vec2-large_gender=male</th>\n",
       "      <th>sup_gender=female</th>\n",
       "      <th>sup_gender=male</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>action=increase, loc=none, n_words=low, trim_dur=low</td>\n",
       "      <td>68.29</td>\n",
       "      <td>88.62</td>\n",
       "      <td>73.48</td>\n",
       "      <td>81.82</td>\n",
       "      <td>123.0</td>\n",
       "      <td>132.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>action=increase, loc=none, dur=low</td>\n",
       "      <td>74.19</td>\n",
       "      <td>93.55</td>\n",
       "      <td>78.74</td>\n",
       "      <td>89.08</td>\n",
       "      <td>124.0</td>\n",
       "      <td>174.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>=</td>\n",
       "      <td>action=increase, n_words=low, speakRate=high</td>\n",
       "      <td>78.41</td>\n",
       "      <td>90.91</td>\n",
       "      <td>75.41</td>\n",
       "      <td>75.41</td>\n",
       "      <td>176.0</td>\n",
       "      <td>122.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>=</td>\n",
       "      <td>action=increase, n_words=low, object=volume</td>\n",
       "      <td>73.44</td>\n",
       "      <td>89.84</td>\n",
       "      <td>78.15</td>\n",
       "      <td>78.15</td>\n",
       "      <td>128.0</td>\n",
       "      <td>119.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>=</td>\n",
       "      <td>speakRate=high, dur=low, tot_silence=medium</td>\n",
       "      <td>85.71</td>\n",
       "      <td>95.65</td>\n",
       "      <td>86.86</td>\n",
       "      <td>86.86</td>\n",
       "      <td>161.0</td>\n",
       "      <td>137.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>=</td>\n",
       "      <td>loc=none, speakRate=high, tot_silence=medium</td>\n",
       "      <td>90.08</td>\n",
       "      <td>95.80</td>\n",
       "      <td>87.16</td>\n",
       "      <td>87.16</td>\n",
       "      <td>262.0</td>\n",
       "      <td>148.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>=</td>\n",
       "      <td>n_words=low, dur=low, tot_silence=medium</td>\n",
       "      <td>85.55</td>\n",
       "      <td>95.38</td>\n",
       "      <td>87.34</td>\n",
       "      <td>87.34</td>\n",
       "      <td>173.0</td>\n",
       "      <td>158.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>=</td>\n",
       "      <td>loc=none, speakRate_trim=medium, dur=low</td>\n",
       "      <td>84.13</td>\n",
       "      <td>94.44</td>\n",
       "      <td>88.33</td>\n",
       "      <td>88.33</td>\n",
       "      <td>126.0</td>\n",
       "      <td>180.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>=</td>\n",
       "      <td>speakRate=medium, speakRate_trim=low</td>\n",
       "      <td>91.82</td>\n",
       "      <td>94.34</td>\n",
       "      <td>90.05</td>\n",
       "      <td>90.05</td>\n",
       "      <td>159.0</td>\n",
       "      <td>211.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>=</td>\n",
       "      <td>action=decrease, n_words=low</td>\n",
       "      <td>91.37</td>\n",
       "      <td>91.37</td>\n",
       "      <td>82.90</td>\n",
       "      <td>92.75</td>\n",
       "      <td>197.0</td>\n",
       "      <td>193.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>action=activate, n_words=medium</td>\n",
       "      <td>96.17</td>\n",
       "      <td>95.22</td>\n",
       "      <td>94.54</td>\n",
       "      <td>78.14</td>\n",
       "      <td>209.0</td>\n",
       "      <td>183.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>action=activate, trim_dur=medium</td>\n",
       "      <td>97.71</td>\n",
       "      <td>94.66</td>\n",
       "      <td>95.42</td>\n",
       "      <td>79.39</td>\n",
       "      <td>131.0</td>\n",
       "      <td>131.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>loc=washroom, object=lights</td>\n",
       "      <td>96.58</td>\n",
       "      <td>96.58</td>\n",
       "      <td>95.65</td>\n",
       "      <td>80.00</td>\n",
       "      <td>117.0</td>\n",
       "      <td>115.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>action=activate, n_words=medium, object=lights</td>\n",
       "      <td>97.63</td>\n",
       "      <td>95.27</td>\n",
       "      <td>94.08</td>\n",
       "      <td>78.95</td>\n",
       "      <td>169.0</td>\n",
       "      <td>152.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Gain                                          Subgroup_str  \\\n",
       "0    >  action=increase, loc=none, n_words=low, trim_dur=low   \n",
       "1    >                    action=increase, loc=none, dur=low   \n",
       "0    =          action=increase, n_words=low, speakRate=high   \n",
       "1    =           action=increase, n_words=low, object=volume   \n",
       "2    =           speakRate=high, dur=low, tot_silence=medium   \n",
       "3    =          loc=none, speakRate=high, tot_silence=medium   \n",
       "4    =              n_words=low, dur=low, tot_silence=medium   \n",
       "5    =              loc=none, speakRate_trim=medium, dur=low   \n",
       "6    =                  speakRate=medium, speakRate_trim=low   \n",
       "7    =                          action=decrease, n_words=low   \n",
       "0    <                       action=activate, n_words=medium   \n",
       "1    <                      action=activate, trim_dur=medium   \n",
       "2    <                           loc=washroom, object=lights   \n",
       "3    <        action=activate, n_words=medium, object=lights   \n",
       "\n",
       "   wav2vec2-base_gender=female  wav2vec2-large_gender=female  \\\n",
       "0                        68.29                         88.62   \n",
       "1                        74.19                         93.55   \n",
       "0                        78.41                         90.91   \n",
       "1                        73.44                         89.84   \n",
       "2                        85.71                         95.65   \n",
       "3                        90.08                         95.80   \n",
       "4                        85.55                         95.38   \n",
       "5                        84.13                         94.44   \n",
       "6                        91.82                         94.34   \n",
       "7                        91.37                         91.37   \n",
       "0                        96.17                         95.22   \n",
       "1                        97.71                         94.66   \n",
       "2                        96.58                         96.58   \n",
       "3                        97.63                         95.27   \n",
       "\n",
       "   wav2vec2-base_gender=male  wav2vec2-large_gender=male  sup_gender=female  \\\n",
       "0                      73.48                       81.82              123.0   \n",
       "1                      78.74                       89.08              124.0   \n",
       "0                      75.41                       75.41              176.0   \n",
       "1                      78.15                       78.15              128.0   \n",
       "2                      86.86                       86.86              161.0   \n",
       "3                      87.16                       87.16              262.0   \n",
       "4                      87.34                       87.34              173.0   \n",
       "5                      88.33                       88.33              126.0   \n",
       "6                      90.05                       90.05              159.0   \n",
       "7                      82.90                       92.75              197.0   \n",
       "0                      94.54                       78.14              209.0   \n",
       "1                      95.42                       79.39              131.0   \n",
       "2                      95.65                       80.00              117.0   \n",
       "3                      94.08                       78.95              169.0   \n",
       "\n",
       "   sup_gender=male  \n",
       "0            132.0  \n",
       "1            174.0  \n",
       "0            122.0  \n",
       "1            119.0  \n",
       "2            137.0  \n",
       "3            148.0  \n",
       "4            158.0  \n",
       "5            180.0  \n",
       "6            211.0  \n",
       "7            193.0  \n",
       "0            183.0  \n",
       "1            131.0  \n",
       "2            115.0  \n",
       "3            152.0  "
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Select columns for better visualization\n",
    "comparison_res_pr = comparison_res.copy()\n",
    "\n",
    "comparison_res_pr['Subgroup_str'] = comparison_res_pr['Subgroup'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "\n",
    "comparison_res_pr = comparison_res_pr[['Gain', 'Subgroup_str', \n",
    "       'wav2vec2-base_gender=female', 'wav2vec2-large_gender=female',\n",
    "       'wav2vec2-base_gender=male', 'wav2vec2-large_gender=male', \n",
    "       'sup_gender=female', 'sup_gender=male']]\n",
    "\n",
    "comparison_res_pr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "748ab973",
   "metadata": {},
   "outputs": [],
   "source": [
    "comparison_res_pr['wav2vec2-base_performance_diff'] = comparison_res_pr['wav2vec2-base_gender=female'] - comparison_res_pr['wav2vec2-base_gender=male']\n",
    "comparison_res_pr['wav2vec2-large_performance_diff'] = comparison_res_pr['wav2vec2-large_gender=female'] - comparison_res_pr['wav2vec2-large_gender=male']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "02d60f3a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Gain</th>\n",
       "      <th>Subgroup_str</th>\n",
       "      <th>wav2vec2-base_gender=female</th>\n",
       "      <th>wav2vec2-large_gender=female</th>\n",
       "      <th>wav2vec2-base_gender=male</th>\n",
       "      <th>wav2vec2-large_gender=male</th>\n",
       "      <th>sup_gender=female</th>\n",
       "      <th>sup_gender=male</th>\n",
       "      <th>wav2vec2-base_performance_diff</th>\n",
       "      <th>wav2vec2-large_performance_diff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>action=activate, n_words=medium</td>\n",
       "      <td>96.17</td>\n",
       "      <td>95.22</td>\n",
       "      <td>94.54</td>\n",
       "      <td>78.14</td>\n",
       "      <td>209.0</td>\n",
       "      <td>183.0</td>\n",
       "      <td>1.63</td>\n",
       "      <td>17.08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>loc=washroom, object=lights</td>\n",
       "      <td>96.58</td>\n",
       "      <td>96.58</td>\n",
       "      <td>95.65</td>\n",
       "      <td>80.00</td>\n",
       "      <td>117.0</td>\n",
       "      <td>115.0</td>\n",
       "      <td>0.93</td>\n",
       "      <td>16.58</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>action=activate, n_words=medium, object=lights</td>\n",
       "      <td>97.63</td>\n",
       "      <td>95.27</td>\n",
       "      <td>94.08</td>\n",
       "      <td>78.95</td>\n",
       "      <td>169.0</td>\n",
       "      <td>152.0</td>\n",
       "      <td>3.55</td>\n",
       "      <td>16.32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>=</td>\n",
       "      <td>action=increase, n_words=low, speakRate=high</td>\n",
       "      <td>78.41</td>\n",
       "      <td>90.91</td>\n",
       "      <td>75.41</td>\n",
       "      <td>75.41</td>\n",
       "      <td>176.0</td>\n",
       "      <td>122.0</td>\n",
       "      <td>3.00</td>\n",
       "      <td>15.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>action=activate, trim_dur=medium</td>\n",
       "      <td>97.71</td>\n",
       "      <td>94.66</td>\n",
       "      <td>95.42</td>\n",
       "      <td>79.39</td>\n",
       "      <td>131.0</td>\n",
       "      <td>131.0</td>\n",
       "      <td>2.29</td>\n",
       "      <td>15.27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>=</td>\n",
       "      <td>action=increase, n_words=low, object=volume</td>\n",
       "      <td>73.44</td>\n",
       "      <td>89.84</td>\n",
       "      <td>78.15</td>\n",
       "      <td>78.15</td>\n",
       "      <td>128.0</td>\n",
       "      <td>119.0</td>\n",
       "      <td>-4.71</td>\n",
       "      <td>11.69</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>=</td>\n",
       "      <td>speakRate=high, dur=low, tot_silence=medium</td>\n",
       "      <td>85.71</td>\n",
       "      <td>95.65</td>\n",
       "      <td>86.86</td>\n",
       "      <td>86.86</td>\n",
       "      <td>161.0</td>\n",
       "      <td>137.0</td>\n",
       "      <td>-1.15</td>\n",
       "      <td>8.79</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>=</td>\n",
       "      <td>loc=none, speakRate=high, tot_silence=medium</td>\n",
       "      <td>90.08</td>\n",
       "      <td>95.80</td>\n",
       "      <td>87.16</td>\n",
       "      <td>87.16</td>\n",
       "      <td>262.0</td>\n",
       "      <td>148.0</td>\n",
       "      <td>2.92</td>\n",
       "      <td>8.64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>=</td>\n",
       "      <td>n_words=low, dur=low, tot_silence=medium</td>\n",
       "      <td>85.55</td>\n",
       "      <td>95.38</td>\n",
       "      <td>87.34</td>\n",
       "      <td>87.34</td>\n",
       "      <td>173.0</td>\n",
       "      <td>158.0</td>\n",
       "      <td>-1.79</td>\n",
       "      <td>8.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>action=increase, loc=none, n_words=low, trim_dur=low</td>\n",
       "      <td>68.29</td>\n",
       "      <td>88.62</td>\n",
       "      <td>73.48</td>\n",
       "      <td>81.82</td>\n",
       "      <td>123.0</td>\n",
       "      <td>132.0</td>\n",
       "      <td>-5.19</td>\n",
       "      <td>6.80</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>=</td>\n",
       "      <td>loc=none, speakRate_trim=medium, dur=low</td>\n",
       "      <td>84.13</td>\n",
       "      <td>94.44</td>\n",
       "      <td>88.33</td>\n",
       "      <td>88.33</td>\n",
       "      <td>126.0</td>\n",
       "      <td>180.0</td>\n",
       "      <td>-4.20</td>\n",
       "      <td>6.11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>action=increase, loc=none, dur=low</td>\n",
       "      <td>74.19</td>\n",
       "      <td>93.55</td>\n",
       "      <td>78.74</td>\n",
       "      <td>89.08</td>\n",
       "      <td>124.0</td>\n",
       "      <td>174.0</td>\n",
       "      <td>-4.55</td>\n",
       "      <td>4.47</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>=</td>\n",
       "      <td>speakRate=medium, speakRate_trim=low</td>\n",
       "      <td>91.82</td>\n",
       "      <td>94.34</td>\n",
       "      <td>90.05</td>\n",
       "      <td>90.05</td>\n",
       "      <td>159.0</td>\n",
       "      <td>211.0</td>\n",
       "      <td>1.77</td>\n",
       "      <td>4.29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>=</td>\n",
       "      <td>action=decrease, n_words=low</td>\n",
       "      <td>91.37</td>\n",
       "      <td>91.37</td>\n",
       "      <td>82.90</td>\n",
       "      <td>92.75</td>\n",
       "      <td>197.0</td>\n",
       "      <td>193.0</td>\n",
       "      <td>8.47</td>\n",
       "      <td>-1.38</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Gain                                          Subgroup_str  \\\n",
       "0    <                       action=activate, n_words=medium   \n",
       "2    <                           loc=washroom, object=lights   \n",
       "3    <        action=activate, n_words=medium, object=lights   \n",
       "0    =          action=increase, n_words=low, speakRate=high   \n",
       "1    <                      action=activate, trim_dur=medium   \n",
       "1    =           action=increase, n_words=low, object=volume   \n",
       "2    =           speakRate=high, dur=low, tot_silence=medium   \n",
       "3    =          loc=none, speakRate=high, tot_silence=medium   \n",
       "4    =              n_words=low, dur=low, tot_silence=medium   \n",
       "0    >  action=increase, loc=none, n_words=low, trim_dur=low   \n",
       "5    =              loc=none, speakRate_trim=medium, dur=low   \n",
       "1    >                    action=increase, loc=none, dur=low   \n",
       "6    =                  speakRate=medium, speakRate_trim=low   \n",
       "7    =                          action=decrease, n_words=low   \n",
       "\n",
       "   wav2vec2-base_gender=female  wav2vec2-large_gender=female  \\\n",
       "0                        96.17                         95.22   \n",
       "2                        96.58                         96.58   \n",
       "3                        97.63                         95.27   \n",
       "0                        78.41                         90.91   \n",
       "1                        97.71                         94.66   \n",
       "1                        73.44                         89.84   \n",
       "2                        85.71                         95.65   \n",
       "3                        90.08                         95.80   \n",
       "4                        85.55                         95.38   \n",
       "0                        68.29                         88.62   \n",
       "5                        84.13                         94.44   \n",
       "1                        74.19                         93.55   \n",
       "6                        91.82                         94.34   \n",
       "7                        91.37                         91.37   \n",
       "\n",
       "   wav2vec2-base_gender=male  wav2vec2-large_gender=male  sup_gender=female  \\\n",
       "0                      94.54                       78.14              209.0   \n",
       "2                      95.65                       80.00              117.0   \n",
       "3                      94.08                       78.95              169.0   \n",
       "0                      75.41                       75.41              176.0   \n",
       "1                      95.42                       79.39              131.0   \n",
       "1                      78.15                       78.15              128.0   \n",
       "2                      86.86                       86.86              161.0   \n",
       "3                      87.16                       87.16              262.0   \n",
       "4                      87.34                       87.34              173.0   \n",
       "0                      73.48                       81.82              123.0   \n",
       "5                      88.33                       88.33              126.0   \n",
       "1                      78.74                       89.08              124.0   \n",
       "6                      90.05                       90.05              159.0   \n",
       "7                      82.90                       92.75              197.0   \n",
       "\n",
       "   sup_gender=male  wav2vec2-base_performance_diff  \\\n",
       "0            183.0                            1.63   \n",
       "2            115.0                            0.93   \n",
       "3            152.0                            3.55   \n",
       "0            122.0                            3.00   \n",
       "1            131.0                            2.29   \n",
       "1            119.0                           -4.71   \n",
       "2            137.0                           -1.15   \n",
       "3            148.0                            2.92   \n",
       "4            158.0                           -1.79   \n",
       "0            132.0                           -5.19   \n",
       "5            180.0                           -4.20   \n",
       "1            174.0                           -4.55   \n",
       "6            211.0                            1.77   \n",
       "7            193.0                            8.47   \n",
       "\n",
       "   wav2vec2-large_performance_diff  \n",
       "0                            17.08  \n",
       "2                            16.58  \n",
       "3                            16.32  \n",
       "0                            15.50  \n",
       "1                            15.27  \n",
       "1                            11.69  \n",
       "2                             8.79  \n",
       "3                             8.64  \n",
       "4                             8.04  \n",
       "0                             6.80  \n",
       "5                             6.11  \n",
       "1                             4.47  \n",
       "6                             4.29  \n",
       "7                            -1.38  "
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "comparison_res_pr.sort_values('wav2vec2-large_performance_diff', ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "8639411c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Gain</th>\n",
       "      <th>Subgroup_str</th>\n",
       "      <th>wav2vec2-base_gender=female</th>\n",
       "      <th>wav2vec2-large_gender=female</th>\n",
       "      <th>wav2vec2-base_gender=male</th>\n",
       "      <th>wav2vec2-large_gender=male</th>\n",
       "      <th>sup_gender=female</th>\n",
       "      <th>sup_gender=male</th>\n",
       "      <th>wav2vec2-base_performance_diff</th>\n",
       "      <th>wav2vec2-large_performance_diff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>action=increase, loc=none, n_words=low, trim_dur=low</td>\n",
       "      <td>68.29</td>\n",
       "      <td>88.62</td>\n",
       "      <td>73.48</td>\n",
       "      <td>81.82</td>\n",
       "      <td>123.0</td>\n",
       "      <td>132.0</td>\n",
       "      <td>-5.19</td>\n",
       "      <td>6.80</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>=</td>\n",
       "      <td>action=increase, n_words=low, speakRate=high</td>\n",
       "      <td>78.41</td>\n",
       "      <td>90.91</td>\n",
       "      <td>75.41</td>\n",
       "      <td>75.41</td>\n",
       "      <td>176.0</td>\n",
       "      <td>122.0</td>\n",
       "      <td>3.00</td>\n",
       "      <td>15.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>action=activate, n_words=medium</td>\n",
       "      <td>96.17</td>\n",
       "      <td>95.22</td>\n",
       "      <td>94.54</td>\n",
       "      <td>78.14</td>\n",
       "      <td>209.0</td>\n",
       "      <td>183.0</td>\n",
       "      <td>1.63</td>\n",
       "      <td>17.08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Gain                                          Subgroup_str  \\\n",
       "0    >  action=increase, loc=none, n_words=low, trim_dur=low   \n",
       "0    =          action=increase, n_words=low, speakRate=high   \n",
       "0    <                       action=activate, n_words=medium   \n",
       "\n",
       "   wav2vec2-base_gender=female  wav2vec2-large_gender=female  \\\n",
       "0                        68.29                         88.62   \n",
       "0                        78.41                         90.91   \n",
       "0                        96.17                         95.22   \n",
       "\n",
       "   wav2vec2-base_gender=male  wav2vec2-large_gender=male  sup_gender=female  \\\n",
       "0                      73.48                       81.82              123.0   \n",
       "0                      75.41                       75.41              176.0   \n",
       "0                      94.54                       78.14              209.0   \n",
       "\n",
       "   sup_gender=male  wav2vec2-base_performance_diff  \\\n",
       "0            132.0                           -5.19   \n",
       "0            122.0                            3.00   \n",
       "0            183.0                            1.63   \n",
       "\n",
       "   wav2vec2-large_performance_diff  \n",
       "0                             6.80  \n",
       "0                            15.50  \n",
       "0                            17.08  "
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Choose the top itemset for each gain value (i.e., >, <, =)\n",
    "comparison_top_1 = comparison_res_pr.loc[[0]]\n",
    "comparison_top_1 = comparison_top_1.sort_values(\"Gain\", ascending=False)\n",
    "comparison_top_1"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "327a8736",
   "metadata": {},
   "source": [
    "#### Shapley value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "eee769bd",
   "metadata": {},
   "outputs": [],
   "source": [
    "height = 0.45\n",
    "sizeFig = (6.5,2.5)\n",
    "labelsize = 13.3\n",
    "pad = 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "4e3b862a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 650x250 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Plot for paper \n",
    "itemset = comparison_res.loc[0].iloc[0]['Subgroup']\n",
    "\n",
    "itemset_shaps = []\n",
    "for gender in genders:\n",
    "    itemset_0 = itemset.union([gender])\n",
    "    itemset_shaps.append(fp_divergence_difference.computeShapleyValue(itemset_0))\n",
    "\n",
    "# Sort \"gender = female\"\n",
    "sorted_dict_1 = {}\n",
    "tmp = {}\n",
    "for key,value in itemset_shaps[0].items():\n",
    "    if \"male\" in str(key):\n",
    "        sorted_dict_1[key] = value\n",
    "    else:\n",
    "        tmp[key] = value\n",
    "for key,value in tmp.items():\n",
    "    sorted_dict_1[key] = value\n",
    "\n",
    "# Sort \"gender = male\"\n",
    "sorted_dict_2 = {}\n",
    "tmp = {}\n",
    "for key,value in itemset_shaps[1].items():\n",
    "    if \"male\" in str(key):\n",
    "        sorted_dict_2[key] = value\n",
    "    else:\n",
    "        tmp[key] = value\n",
    "for key,value in tmp.items():\n",
    "    sorted_dict_2[key] = value\n",
    "\n",
    "name_fig = f\"plot_results/comparison_shapley_w2b_w2l\"\n",
    "        \n",
    "plotComparisonShapleyValues(abbreviateDict(sorted_dict_1, abbreviations), \\\n",
    "                            abbreviateDict(sorted_dict_2, abbreviations), \\\n",
    "                            sizeFig=sizeFig, sharedAxis=True, sort=\"paper\", \\\n",
    "                            height = height, saveFig=True, subcaption=False, \\\n",
    "                            nameFig=name_fig, labelsize=labelsize, pad=pad)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "f3be0778",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 650x250 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Plot Shapley Values for the first just-retrieved data subgroup\n",
    "itemset = comparison_res.loc[0].iloc[0]['Subgroup']\n",
    "\n",
    "itemset_shaps = []\n",
    "for gender in genders:\n",
    "    itemset_0 = itemset.union([gender])\n",
    "    itemset_shaps.append(fp_divergence_difference.computeShapleyValue(itemset_0))\n",
    "                \n",
    "plotComparisonShapleyValues(abbreviateDict(itemset_shaps[0], abbreviations), \\\n",
    "                            abbreviateDict(itemset_shaps[1], abbreviations), \\\n",
    "                            sizeFig=sizeFig, sharedAxis=True, sort=\"alp\", \\\n",
    "                            height = height, saveFig=False, subcaption=False, \\\n",
    "                            labelsize=labelsize, pad=pad) "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e6463273",
   "metadata": {},
   "source": [
    "## Global Shapley value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "3f110edc",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Compute the top-K global shapley values related to the gain in performance between Wav2Vec2-Base and Wav2Vec2-Large\n",
    "global_item_divergence_w2b_w2l = fp_divergence_difference.computeGlobalShapleyValue()\n",
    "\n",
    "K = 15\n",
    "topK_global_w2b_w2l = {k:v for k,v in global_item_divergence_w2b_w2l.items() \\\n",
    "                        if k in sorted(global_item_divergence_w2b_w2l, \n",
    "                        key=lambda x: abs(global_item_divergence_w2b_w2l[x]))[::-1][:K]}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "8a635eba",
   "metadata": {},
   "outputs": [],
   "source": [
    "from utils_analysis import plotShapleyValue\n",
    "\n",
    "sizeFig = (3.2,4)\n",
    "labelsize = 13\n",
    "titlesize = 13"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "d2f0a189",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 320x400 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Plot and Save the image \n",
    "topK_global_w2b_w2l_abbr = abbreviateDict(topK_global_w2b_w2l, abbreviations )\n",
    "\n",
    "plotShapleyValue(shapley_values=topK_global_w2b_w2l_abbr, \\\n",
    "                sizeFig=sizeFig, labelsize=labelsize, titlesize=titlesize, \\\n",
    "                title = r\"$\\tilde{\\Delta}^g_{gain} Wav2B - Wav2L$\", saveFig=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ba25022",
   "metadata": {},
   "source": [
    "# Divergence difference Wav2Vec2-Base vs Hubert-Base"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "dc56bafe",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Compute the divergence for Hubert-Base\n",
    "config = 'hubert-base'\n",
    "fp_divergence_i = fp_divergence_dict[config]\n",
    "FPdiv_hb = fp_divergence_i.getDivergence(th_redundancy=None).copy()\n",
    "hbase = FPdiv_hb.set_index(\"itemsets\")\n",
    "\n",
    "## Compute the divergence for Wav2Vec2-Base\n",
    "config = 'wav2vec2-base'\n",
    "fp_divergence_i = fp_divergence_dict[config]\n",
    "FPdiv_w2b = fp_divergence_i.getDivergence(th_redundancy=None).copy()\n",
    "wbase = FPdiv_w2b.set_index(\"itemsets\")\n",
    "\n",
    "## Merge the df\n",
    "merged = wbase.join(hbase, lsuffix='_w2b', rsuffix='_hb')\n",
    "merged = merged.rename(columns={'support_hb': 'support'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "32256e51",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Compute difference in performance between the models\n",
    "diff = \"d_difference\"\n",
    "merged[diff] = merged[\"accuracy_hb\"] - merged[\"accuracy_w2b\"]\n",
    "merged[\"difference\"] = merged[\"accuracy_hb\"] - merged[\"accuracy_w2b\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "2c799c4d",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Create \"Gain W2Base-HBase\" df and compute divergence\n",
    "hbase_w2base_gain_df = merged[['support', '#errors_hb', '#corrects_hb', 'accuracy_hb',\n",
    "       'd_accuracy_hb', 't_value_hb', 'support_count_hb', 'length_hb'] \\\n",
    "       + [diff, \"difference\", \"accuracy_w2b\"]]\n",
    "hbase_w2base_gain_df = hbase_w2base_gain_df.rename(columns={'length_hb':'length'})\n",
    "hbase_w2base_gain_df = hbase_w2base_gain_df.reset_index()\n",
    "\n",
    "fp_divergence_difference = FP_Divergence(hbase_w2base_gain_df, diff)\n",
    "diff_nr = fp_divergence_difference.getDivergence(th_redundancy=0.0) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "5981f5a4",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Select meaningful columns \n",
    "sel = diff_nr.itemsets.values\n",
    "compare_performance = merged.loc[sel].sort_values(diff, ascending = False)\n",
    "cols = ['d_difference', 'accuracy_w2b', 'accuracy_hb', 'support']"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c7560518",
   "metadata": {},
   "source": [
    "## Stats"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "a76357a3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Equal\n",
      "1.927\n",
      "Greater\n",
      "97.034\n",
      "Lower\n",
      "1.039\n"
     ]
    }
   ],
   "source": [
    "## Compute meaningful statistics:\n",
    "diff_nr_0 = fp_divergence_difference.getDivergence(th_redundancy=None)\n",
    "\n",
    "# Percentage of itemsets for which performance are equal for the two model\n",
    "print(\"Equal\")\n",
    "print(round(100*(diff_nr_0.loc[(diff_nr_0[diff])==0].shape[0]/diff_nr_0.shape[0]),3))\n",
    "\n",
    "# Percentage of itemsets for which performance are greater for Hubert-Base\n",
    "print(\"Greater\")\n",
    "print(round(100*(diff_nr_0.loc[(diff_nr_0[diff])>0].shape[0]/diff_nr_0.shape[0]), 3))\n",
    "\n",
    "# Percentage of itemsets for which performance are lower for Hubert-Base\n",
    "print(\"Lower\")\n",
    "print(round(100*(diff_nr_0.loc[(diff_nr_0[diff])<0].shape[0]/diff_nr_0.shape[0]),3))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "01d0c446",
   "metadata": {},
   "source": [
    "## Gain > 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "3d4c0e17",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_w2b</th>\n",
       "      <th>accuracy_hb</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>(location=none, gender=male, trimmed_duration=low, n_words=low, total_silence=high)</td>\n",
       "      <td>31.20</td>\n",
       "      <td>64.00</td>\n",
       "      <td>95.20</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>(gender=male, trimmed_duration=low, ageRange=22-40, object=volume, n_words=low)</td>\n",
       "      <td>31.11</td>\n",
       "      <td>67.41</td>\n",
       "      <td>98.52</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>(location=none, gender=male, speed_rate_word=high, trimmed_duration=low, n_words=low, total_silence=high)</td>\n",
       "      <td>30.89</td>\n",
       "      <td>64.23</td>\n",
       "      <td>95.12</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>(ageRange=22-40, object=volume, speed_rate_word=high, gender=male)</td>\n",
       "      <td>30.77</td>\n",
       "      <td>67.52</td>\n",
       "      <td>98.29</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>(n_words=low, trimmed_duration=low, total_silence=high, gender=male)</td>\n",
       "      <td>30.77</td>\n",
       "      <td>65.04</td>\n",
       "      <td>95.80</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>(gender=male, speed_rate_word=high, trimmed_duration=low, n_words=low, total_silence=high)</td>\n",
       "      <td>30.50</td>\n",
       "      <td>65.25</td>\n",
       "      <td>95.74</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>(ageRange=22-40, object=volume, trimmed_duration=low, gender=male)</td>\n",
       "      <td>30.19</td>\n",
       "      <td>68.55</td>\n",
       "      <td>98.74</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>(speed_rate_word_trimmed=high, speed_rate_word=high, object=volume, n_words=low, action=increase)</td>\n",
       "      <td>30.17</td>\n",
       "      <td>68.10</td>\n",
       "      <td>98.28</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>(gender=male, speed_rate_word=high, ageRange=22-40, location=none, total_silence=high)</td>\n",
       "      <td>29.41</td>\n",
       "      <td>60.50</td>\n",
       "      <td>89.92</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>(gender=male, speed_rate_word=high, trimmed_duration=low, location=none, total_silence=high)</td>\n",
       "      <td>29.32</td>\n",
       "      <td>64.66</td>\n",
       "      <td>93.98</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>(object=volume, speed_rate_word_trimmed=high, speed_rate_word=high, action=increase)</td>\n",
       "      <td>29.27</td>\n",
       "      <td>69.11</td>\n",
       "      <td>98.37</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>(speed_rate_word=high, trimmed_duration=low, location=none, action=increase, total_duration=low)</td>\n",
       "      <td>29.17</td>\n",
       "      <td>68.33</td>\n",
       "      <td>97.50</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>(total_silence=high, trimmed_duration=low, speed_rate_word=high, gender=male)</td>\n",
       "      <td>29.14</td>\n",
       "      <td>65.56</td>\n",
       "      <td>94.70</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>(speed_rate_word=high, trimmed_duration=low, ageRange=22-40, object=volume, n_words=low)</td>\n",
       "      <td>28.99</td>\n",
       "      <td>66.86</td>\n",
       "      <td>95.86</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>(location=none, gender=male, speed_rate_word=high, trimmed_duration=low, ageRange=22-40, n_words=low)</td>\n",
       "      <td>28.99</td>\n",
       "      <td>67.63</td>\n",
       "      <td>96.62</td>\n",
       "      <td>0.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>(ageRange=22-40, object=volume, speed_rate_word=high, trimmed_duration=low)</td>\n",
       "      <td>28.82</td>\n",
       "      <td>67.06</td>\n",
       "      <td>95.88</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>(gender=male, speed_rate_word=high, trimmed_duration=low, ageRange=22-40, n_words=low)</td>\n",
       "      <td>28.70</td>\n",
       "      <td>68.16</td>\n",
       "      <td>96.86</td>\n",
       "      <td>0.06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>(ageRange=22-40, total_silence=high, speed_rate_word=high, gender=male)</td>\n",
       "      <td>28.57</td>\n",
       "      <td>62.86</td>\n",
       "      <td>91.43</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>(gender=male, speed_rate_word=high, trimmed_duration=low, ageRange=22-40, location=none)</td>\n",
       "      <td>28.37</td>\n",
       "      <td>67.44</td>\n",
       "      <td>95.81</td>\n",
       "      <td>0.06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>(ageRange=22-40, trimmed_duration=low, speed_rate_word=high, gender=male)</td>\n",
       "      <td>28.14</td>\n",
       "      <td>67.96</td>\n",
       "      <td>96.10</td>\n",
       "      <td>0.06</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                     itemsets  \\\n",
       "0                         (location=none, gender=male, trimmed_duration=low, n_words=low, total_silence=high)   \n",
       "1                             (gender=male, trimmed_duration=low, ageRange=22-40, object=volume, n_words=low)   \n",
       "2   (location=none, gender=male, speed_rate_word=high, trimmed_duration=low, n_words=low, total_silence=high)   \n",
       "3                                          (ageRange=22-40, object=volume, speed_rate_word=high, gender=male)   \n",
       "4                                        (n_words=low, trimmed_duration=low, total_silence=high, gender=male)   \n",
       "5                  (gender=male, speed_rate_word=high, trimmed_duration=low, n_words=low, total_silence=high)   \n",
       "6                                          (ageRange=22-40, object=volume, trimmed_duration=low, gender=male)   \n",
       "7           (speed_rate_word_trimmed=high, speed_rate_word=high, object=volume, n_words=low, action=increase)   \n",
       "8                      (gender=male, speed_rate_word=high, ageRange=22-40, location=none, total_silence=high)   \n",
       "9                (gender=male, speed_rate_word=high, trimmed_duration=low, location=none, total_silence=high)   \n",
       "10                       (object=volume, speed_rate_word_trimmed=high, speed_rate_word=high, action=increase)   \n",
       "11           (speed_rate_word=high, trimmed_duration=low, location=none, action=increase, total_duration=low)   \n",
       "12                              (total_silence=high, trimmed_duration=low, speed_rate_word=high, gender=male)   \n",
       "13                   (speed_rate_word=high, trimmed_duration=low, ageRange=22-40, object=volume, n_words=low)   \n",
       "14      (location=none, gender=male, speed_rate_word=high, trimmed_duration=low, ageRange=22-40, n_words=low)   \n",
       "15                                (ageRange=22-40, object=volume, speed_rate_word=high, trimmed_duration=low)   \n",
       "16                     (gender=male, speed_rate_word=high, trimmed_duration=low, ageRange=22-40, n_words=low)   \n",
       "17                                    (ageRange=22-40, total_silence=high, speed_rate_word=high, gender=male)   \n",
       "18                   (gender=male, speed_rate_word=high, trimmed_duration=low, ageRange=22-40, location=none)   \n",
       "19                                  (ageRange=22-40, trimmed_duration=low, speed_rate_word=high, gender=male)   \n",
       "\n",
       "    d_difference  accuracy_w2b  accuracy_hb  support  \n",
       "0          31.20         64.00        95.20     0.03  \n",
       "1          31.11         67.41        98.52     0.04  \n",
       "2          30.89         64.23        95.12     0.03  \n",
       "3          30.77         67.52        98.29     0.03  \n",
       "4          30.77         65.04        95.80     0.04  \n",
       "5          30.50         65.25        95.74     0.04  \n",
       "6          30.19         68.55        98.74     0.04  \n",
       "7          30.17         68.10        98.28     0.03  \n",
       "8          29.41         60.50        89.92     0.03  \n",
       "9          29.32         64.66        93.98     0.04  \n",
       "10         29.27         69.11        98.37     0.03  \n",
       "11         29.17         68.33        97.50     0.03  \n",
       "12         29.14         65.56        94.70     0.04  \n",
       "13         28.99         66.86        95.86     0.04  \n",
       "14         28.99         67.63        96.62     0.05  \n",
       "15         28.82         67.06        95.88     0.04  \n",
       "16         28.70         68.16        96.86     0.06  \n",
       "17         28.57         62.86        91.43     0.04  \n",
       "18         28.37         67.44        95.81     0.06  \n",
       "19         28.14         67.96        96.10     0.06  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Retrieve the data cohorts for which Hubert-Base performs better than Wav2Vec2-Base\n",
    "pr = compare_performance[cols].head(20).reset_index()\n",
    "pr[\"support\"] = pr[\"support\"].round(2)\n",
    "pr[\"accuracy_hb\"] = (pr[\"accuracy_hb\"]*100).round(2)\n",
    "pr[\"accuracy_w2b\"] = (pr[\"accuracy_w2b\"]*100).round(2)\n",
    "pr[\"d_difference\"] = (pr[\"d_difference\"]*100).round(2)\n",
    "\n",
    "display(pr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "86746d6b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_w2b</th>\n",
       "      <th>accuracy_hb</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>gender=male, loc=none, n_words=low, tot_silence=high, trim_dur=low</td>\n",
       "      <td>31.20</td>\n",
       "      <td>64.00</td>\n",
       "      <td>95.20</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ageRange=22-40, gender=male, n_words=low, object=volume, trim_dur=low</td>\n",
       "      <td>31.11</td>\n",
       "      <td>67.41</td>\n",
       "      <td>98.52</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                itemsets  \\\n",
       "0     gender=male, loc=none, n_words=low, tot_silence=high, trim_dur=low   \n",
       "1  ageRange=22-40, gender=male, n_words=low, object=volume, trim_dur=low   \n",
       "\n",
       "   d_difference  accuracy_w2b  accuracy_hb  support  \n",
       "0         31.20         64.00        95.20     0.03  \n",
       "1         31.11         67.41        98.52     0.04  "
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Abbreviate itemset names for better visualization\n",
    "pr_l = pr.head(2).copy()\n",
    "pr_l['itemsets'] = pr_l['itemsets'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "pr_l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "2d4d940c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAATMAAADCCAYAAAAl4SVWAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAeoklEQVR4nO3deZhU1Z3/8feHpVsbqTYOg2CUuCtm4oZLXAhoYtBJxp+JGY0acMuCTiKOGhOcxCULjDtqNEQfDdI/jRGNJhh3Iy64RNyVRXFFRFC2Zl+/88c5BZeiqrq7upuquv19Pc99qu+95557bnXXt88599Y5MjOcc67adSp3AZxzri14MHPOpYIHM+dcKngwc86lggcz51wqeDBzzqWCBzPnXCp0KXcBXPWRJGAbYFG5y+I6hO7Ax9bEQ7EezFwptgE+KnchXIeyLTCzWAIPZq4UiwBmzJhBJpMpd1lcijU2NrLddttBM1oBHsxcyTKZjAczVzH8BoBzLhU8mDnnUsGDmXMuFbzPzJWscelS6OJ/Qq71arp0YbOamlbl4X+JrmS7DxmGunYtdzFcCvSozzBlzKhWBTQPZq5kg4cMpvuWnyt3MVyVW7lyJTfffDMrV6/2YObKo6amhtra2nIXwzmgwm4ASBooySRt2Zo01Sit1+XcplJRwayZngF6AwvbIjMPIs6lQ9U1M81sJfBJucvhnKssm7xmJqlW0rWS5khaLulpSfvnJDtE0qtx//OSvpQ4fqOalKSDJT0paZmkGTH/bjnnvCzuWyHpbUmnS9oeeDwmmx/zHVPCNf1J0h0527pK+kzSqS247uTxF0t6JWfb2ZLeT6yPkXSvpAskzZa0QNJFkrpIulzSPEkfSTotJ5/PS/qzpPmS5kr6a3wvnKta5WhmXgYcC5wM7AtMBx6StFUizeXAecD+wBzgb5LyPgMQA91DwF+APYHjgUOB3yWSjQW+C5wF9AWGAouBGbEsALsRmq/DYr4XSFrcxNI/HnsbcLSkLRLnHAR0A+5uwXWX4nDCKBZfAc4BLgbuA+YDBwKjgdGStovXVUcI4IvjMYfGnx+UlPdWUgzEmexCGJLFuYqySZuZsbZ0BnCKmT0Qt/0AOAI4HXghJr3EzB6J+08mDDfzLeDOPNn+FLjdzEbF9bclnQU8IekMoA9wHHCEmT0a07ybKNO8+OMcM1uQyHd0gfMlZYckeQhYEsvYELedCIw3s8ZmXPflTZynmHnAWWa2Fpgm6XygzsxGxPOMBH4OHALcQQjqa4HvZ8eHirXHBcBA4OE85xgOXNSKMjrX7jZ1n9lOQFdgYnaDma2S9E9CjSkbzJ5N7J8naVrcn08/YGdJJyW2iVDr3AH4ErAGeKIlBTWzeYRA0Zy0qySNA04CGmLw+n+EgAZNX3drvBkDWdZs4I3EedZImgv0jJv6ATsDi8IYi+tsFsuZz0jgqsR6d3w8M1dhNnUwy356ckeMVJ5tuQrt7wT8Abg2z74PCR/cFpN0AXBBE8mOMrOn4s+3EWqDPQk1ruXAA9ns4mtLrntt4risfE3tVTnrVmBbtkuhE/AiIfDm+jRfQcxsBbAiu54TBJ2rCJs6mE0HVhL6aW6H0FEO7AeMSqT7MiEQIelzwK7A1AJ5vgR80cym59sp6XXCB3gA8GieJCvja+ec7S1pZkJ4ZGQGoc/uKGBcvPMKzb/upE+BXpKUGC547ybK0xwvxTLOMbPGNsjPuYqwSYOZmS2R9Hvg8thX9SFwPlAH3AzsFZNeGJtGs4HfAp8B9xbI9lLgOUnXAzcR+q76EvrIfmJm70u6Fbgl9qW9CnwB6GlmdwIfEGou35R0P7DMzBa3pJkZr80k3U64ubArcFgLrjufCcC/AudLugs4khAkWxuAbiP0M/5V0oWE5mIf4NvA5WbmzUdXlcpxN/PnhDt8DYRaws7AIDObn5PmGkJzqDdwdKKWswEze41Q69oFeAp4Gfg1MCuR7AzgLuAGQg3vJsKdRsxsJqFz+38JwTN5F7SlbgP2INTYJubsa851J69rCnAm8F+EAHwAcEUrypbNdynhLuaHhDvAU4BbgM1pfaB0rmzUxIQnFUfSIEJf1GaFApxrX/HxjIVDG8b7F81dq2W/aD5z3I1k6uo22NfY2Eh9fT1AfVPdIlX1DQBJWxPuEr7tgaz8GsY2+BBArk30qM9Q08qx8aoqmAH3Ex4LOLPcBXEwdew1PqGJaxMdbnBGM+tX7jK49TJ1dRs1C5wrl2ocNcM55zbiwcw5lwoezJxzqeDBzDmXCh7MnHOp4MHMOZcKHsycc6ngwcw5lwoezJxzqeDBzDmXCh7MnHOp4MHMOZcKVfVFc1dZGpcuhVYO2+LSoy1GvmgN/0t0Jdt9yDAfz8yt06M+w5Qxo8oW0DyYuZINHjLYR5p1wPrRYleuXu3BzFWfmpoaamtry10M54CU3ACQdLGkVzbxObeXZJL23pTnjec2Scds6vM6V8kqOphJmiBpVDOSXgF8tZ2L45yrYFXdzFSYWruzmS0GFpe7PC0lqauZ5c4+7pwrQcXWzCSNIcyHOSw2q0zSKfF1kKRJwAqgf24zU9IYSfdKukDSbEkLJF0kqYukyyXNk/SRpNNaUJ4DJL0saXk89z45+0+RtCBn2zGSLLF+saRXJJ0m6V1gRQzIrSLpS5L+IWmZpLmSbpS0RWLfWkk94vrn4vq4xPHDJT1bJP9aSZnsQphUxrmKUrHBDBgGPEuYsLd3XGbEfZcBwwkzl79W4PjDgW0IE96eA1wM3AfMBw4ERgOjJW3XVEEkdYvHTgP6xbxKnZB3Z+A44Fhg75j/BZIWN7H0L1C2OuDBeF37A/8JfI31kxm/Acwl/GOA8H7Mja9ZA4EnipR5OLAwsfis567iVGwwM7OFwEpgqZl9YmafAGvi7gvN7BEze8fM5hbIYh5wlplNM7NbCIGozsxGmNnbwMiY/yHNKM5JQGfgNDN708zuAy4v8dJqgMFm9rKZvWZhFubRhMBWbJlUpGybA0PM7A0z+wfwY2CwpK1j/k8SAhbx9Vagk6Q9JHUBDgYmFCnzSKA+sWzbgut1bpOo1j6zQh/spDfNbG1ifTahlgKAma2RNBfo2Yy8+gKvmtnSxLaCzbImfGBmnyY3mNk8QvAtRbZsSxLbJhL+Ue1GuO4JwA/jvgHAL4Ed4s/1hGA4sdAJzGwFoUkPQBu0jJ1rcxVbM2vCkqaTkNuxbgW2Nec9aM6nd22edPkej9+o7K1pZsZzWoF92e0TgC9K2hn4N+ApQrNyAKGm9qKZLSp2cc5Vukqvma0kNO/KbTKh2ba5mS2L276ck+ZToLukbola0t7NzH80cGcTaWYWKdvJOec9hBBc34rr2X6zXxBqcY2SniD0hX2O4v1lzlWFSq+ZvQ8cGB9Q7UH5yns7ITjcHPuZ/h04LyfN88BSYISknSWdCJzSnMzNbJ6ZTW9iWVbg8NuA5cCtkv5N0mHAdUCDmc2O+Wf7zb7H+r6x1wj9d1+leH+Zc1Wh0oPZFYRO/8mEmk+fchQiPsf2H8AewMvAb4Gf5aSZRwgW/w68DpxAuOvZ3mVbCgwCtgJeAO4CHiPcBEh6nFDLnRCPM0JzE+Dp9i6nc+1N4W/aueaLz5otHNow3r9o7oD1XzSfOe5GMnV1bZZvY2Mj9fX1APVm1lgsbaX3mbkK1jC2wYcAcuv0qM9QU8bx7bxmRribCFxQYPdTZnbUpixPpcvWzGbMmkUmkyl3cVyFaI/BGb1m1nLF7iYW6njv8DJ1dW3apHCuNTyY0eqHVp1zFaDS72Y651yzeDBzzqWCBzPnXCp4MHPOpYIHM+dcKngwc86lggcz51wqeDBzzqWCBzPnXCp4MHPOpYJ/ncmVrHHpUijjKAlp1B5f1u4o/C/RlWz3IcN8CKA21qM+w5QxozyglcCDmSvZ4CGDfXDGNpQd4HDl6tUezErgwcyVrKamhtra2nIXwznAbwA451KiIoOZpDGSLC6rJX0o6feSUtGmkTRc0guSFkmaI+leSbsl9neVdKmk1yUtkfSxpLGStmnBOWolvRLfw71z9vWRND7m/ZmkayV5u8ZVtYoMZtGDQG9ge+D7hNmRbihngdrQAOB6wtybRxCa+w9L6hb31wH7Ar+Or98GdgX+1oJzXAZ8nLtRUmfg70A34FDgu8CxwJWlXIhzlaLFwUzSkZKelrRA0lxJ90naKbH/4FgjWC5pkqRjcmsHce7J++NM3bMlNcR5MZNWmNknZvaRmT0M/Bn4eiKPzpJulvSepGWSpkkallPWMbHWc56kWbG810vqmkjTW9LfYx7vSTpR0vuSzk6kqZd0Y6xFNUr6h6S9WvreZZnZkWY2xszeNLNXgVMJ0+j1i/sXmtkRZnanmU0zs+eAnwD9JDU53Z6ko+J7lTu3J3H7HsD3zOxlM3sUOBf4QRzb37mqVErNrBtwFbA/YQLZtcA9kjpJ6g6MJ8wbuS/wS+DS5MGSehNm0H4F2A84EtiaIjN6S9oxpluVU/aPgOMIH85fESbgPS7n8MOAneLryYSJeU9J7B8LbAMMJNRQfgj0TJxbhJpML8KcmP2Al4DHJG0V0/SPgbnYUmjCFID6+Fps6O56wIAFRdIgaWvgJmAwYVLiXAcBb5hZstb2EFAbry1fnrWSMtkF6F6sDM6VQ4vvZprZ3cl1SacDcwgB5VDCB+4HZrYcmCzp84QPV9YZwEtmdkEij9OAGZJ2NbO34uZvSlpMmLh2s7jtnEQ5VgEXJfJ9T9LBhOCWDIzzgR+b2RpgqqS/E4LwTZJ2B74G7G9mk2JZvg+8nTj+MOBLQE8zWxG3nSfpGOA7wI3AJGDvwu8aUCBQxWB5FfC0mb1RIM1mwP8CtxeboSbmNQYYbWaTJG2fJ1kvYHZyg5nNl7Qy7stnOBu+185VnBYHs9ik/DWhv6cH62t3fYDdgNdiIMv6Z04W/YDDYqDKtROQDWaPEwJfHaHPbFfgupyyDI37vgBsDtQQanxJb8ZAljWLEJyI5V1NqGkBYGbTJc3PKe8WwNwQK9bZPJYXM1sGTM9zPc3xO2BPwj+CjcQm8R2E9/nMxPbRhBnUs+XegtAUzQAjmzhnvvkFVWA7Mb+rEuvdCbVi5ypGKc+ZjQdmAD8gdDB3At4gBJJ8HwjlrHeKefwsT96zEj8vMbNsgDhL0uOE2sEvAWJz8mpCf8+zwCLgp8CBOXmuylk31gfg3LLlK3OnWK6BedItiGXpDzxQIK+sEWY2YoOTSNcBRwNfMbONgkMMZHcCOwCH59TKLgSuyDnkcMI/mRU5gXeSpNvM7GTgE3LeI4W7xF3JqbFlxRrpikT6IpfpXHm0KJhJ+hegL/AjM3sqbkvWKKYCJ0mqTTTJ9svJ5iVC39T7Zra6Bae/BHhA0u9jf09/4BkzW3eHM3kjopmmEt6DfYAXYx47A1vmlLcXsNrM3i+QT4uambE5eB3wLWCgmb2XmzgRyHYBDjOzucn9ZjaH0LxPOgv4RWJ9G0J/2PHA83Hbs8D/SOptZtl/Hl8nBKsXm7gG5ypWS28AzAfmAj+UtLOkw9mw+XF7zPNGSX0lDWL9HbVsje16YCvgT5IOkLSjpK9LuiU+NpCXmU0A3mT9zOPTgf0kDZK0q6RfE25KNJuZTQUejeU9QNI+hD6wZYnyPkoIAPfGc20f79j+RtJ+MZ9lZja9iSXZZ3Y9oYl4IrBIUq+4bA4gqQtwF+EfwUlA50Sags+DmdmHZvZGdmF9k/2dRM3vYWAy0CBpH0lfJdTwbmpqxmjnKlmLgpmZrSU8l9SP0LS8mtC0y+5vJDwPtjeh7+q3hLuMAMtjmo+BQwgd+w/FfK4BFhLujBZzFeERgu0Is5D/hfDIxvPAv1Dac2hDCM2rJ4F7CDcrFiXKa4S7mE8CtxACxB2E59/yNsua4QzC3ckJhCZsdjk+7t+W0PzclvA+JtMcXOI5AYj9h98gXN9EQu3vXvI/xuFc1VD4rLbjCaSTgD8C9bGjvKJJ2pbQJ/g1M3us3OWpRPHxjIVDG8b7F83bUPaL5jPH3Uimrq7cxakIjY2N1NfXQ4gfRVsObf5Fc0lDgHeBmcBehOfM7qzUQBabylsQno3rTXhy/n1CTcwV0TC2wYcAamM96jPU+BhxJWmPd60XoWnZi9AsGgf8Tzucp610BUYAOxKal88AJ8Xn2FwRU8deQybjXxpoSz44Y+navZnp0ifbzFy4cKEHM9euWtLMrOQvmjvnXLN5MHPOpYIHM+dcKngwc86lggcz51wqeDBzzqWCBzPnXCp4MHPOpYIHM+dcKngwc86lggcz51wqeDBzzqWCjzXiSta4dCl0sOFqfFSLytWx/hJdm9p9yLAON55Zj/oMU8aM8oBWgTyYuZINHjK4Q400mx0JduXq1R7MKpAHM1eympoaamtry10M5wC/AdBuJI2RdG875f2+pLPbI2/nqpUHM+dcKngwa6U4Wa9zrsyqIphJmiDpWkmXSZon6RNJFzfjuCsljU+sny3JJH0jsW2apB/FnztJulDSR5JWSHpF0pGJtNvH44+LZVoOfE9SZ0lXSVogaa6kywDllOU7kl6XtCymeVRStzZ4e5DUR9JfJS2W1CjpTklbx331ktZI6hfXFd/DFxLHnyBpVpH8ayVlsgvQvS3K7VxbqopgFp0MLAEOBM4HLpR0RBPHTAD6S8pe5wDgs/iKpF7ArsATcf8w4FzChLh7EiYp/pukXXLyvRS4Fugb05wLnAacDhxKmLH9W9nEknoDfyJMItwXGEiYwFhx/0kxEBVbTsp3gZJEmMR3q3hdRwA7ESZHxswWEiYSHhgP2TP7GgMTcV/2PchnOGGS5uzyUZG0zpVFNd3NfM3MLok/vy3px8BXgUeKHPMkoRaxj6SXgP7AFcC34/7DgNlmNjWunwdcamZ3xPWfSToMOBv4r0S+o8zsL9mV2Bk/0szujutDgUGJ9L0J7/VfzOyDuO31xP6/EWZlL6bQ7OlfIwSoHcxsRjz/YOBNSfub2QuEoD4QuDK+PkaYWu9Q4P647eoi5x5JmE0+qzse0FyFqapglrM+C+hZ7AAzWyjpFcKHdRWwFvgDcImk7iRqJLGWsg0wMSebiYTJjJMmZX+QVE8IVs8mzrta0iTWNzVfJQSQ1yU9BDwM3GVm82P6RYQ5O0vRF5iRDWQxv8mSFsR92WB2eqyhDohl+RAYEIN8sna6ETNbAaxIXHOJRXWu/VRTMzN3Ul6jeeWfQAhaA4AnYgB5Ezgkbp+QJ98k5dm2pBnnXZ+h2RpC8+8oYDLwE2CapB2gdc3MAuXL3Z6toe5LqJ1OIASvAYTa6Rwzm9KSa3Ku0lRTzaxUEwh9WauBR+O2J4DvkqiRmFmjpI8JTa8nE8cfDPyzUOax9jcL+HL2OEldgH7AS4l0RqjlTZT0K+ADQr/aVbSumTkZ6CNpu0Qzcw+gHpiSKOMrwI8JAW4y8DGwD/BNiveXOVcVOkIwy9ZK/gP4Rdw2Abgb+NTMJifSXk5ogr5D6DQ/FdgbKFQryroG+LmktwkB5Bxgy+xOSQcS+vceBuYQbmL8K+uDTWuamY8SmuC3xb67LsANhFropES6CYQbHPfEwDpf0mTgeOCsEs/tXMWopmZmSeLdvJeBeYQaCcBThGvPrZFcS+gkv5LQQX8kcLSZvd3Eaa4ExgJjCH1ni4B7Evsbga8QOtvfAn4DnGtmD5R0UQkxMB0DzCcE7keBdwlBKulxoDMbNqufiNu8ZuaqnsJnwbnmizdLFg5tGN8hv2g+c9yNZOrqyl2cDqGxsZH6+nqAejNrLJa2IzQzXTtpGNvQIYcAqulgY7hVi6qumcU7fH8osPsDM/vipixPR5Gtmc2YNYtMJtNk+jTxwRk3rY5UMyt2FzD3UQ7XxjJ1dd7cchWjqoNZK+8COudSJPV3M51zHYMHM+dcKngwc86lggcz51wqeDBzzqWCBzPnXCp4MHPOpYIHM+dcKngwc86lggcz51wqVPXXmVx5NS5dCikZQcK/QF790vGX6Mpi9yHDUjMEUI/6DFPGjPKAVsU8mLmSDR4yOBWDM2YHXVy5erUHsyrmwcyVrKamhtra2nIXwznAbwDkJWmMpHvLXY62krbrcS4fD2bOuVTwYNYOFHgT3rlNqKKDmaTukm6TtETSLEn/LWmCpFFxf42kyyTNjGmelzQwcfwpkhZIGiRpSpwZ/EFJvRNpOku6KqabK+kywmzgyXJI0vmS3pW0TNKrkr6T2D9QksXzTAJWEGYOb+n1TpB0naRRkuZLmi3ph5K6SfqjpEWS3pF0VE75b5b0XizbNEnDmjhP0etxrhpVdDAjzPZ9CHA0cAQhQOyb2P/HuP+7wJ7AOOBBSbsk0tQB5wGDCXNX9gGuSOw/FziNMOv5ocBWhJnGk35DmBD4DOCLwNXA/5c0ICfdZcBwoC/wmqT+MYAWWy7IyeNk4DPgAOA64Pfxup6J1/4Q0CApO/h+J+Aj4DhgD+BXwAhJx230brb8egCQVCspk10Ikyo7V1EqtikkqTvhg32imT0Wt50KfBx/3gk4AdjWzD6Oh10h6UjCBzUbJLoCQ83snXjc74ALE6c6GxhpZnfH/UOBQYlydCPMUH64mT0bN78r6VDgR2w4ge6FZvZI4thJhBnRi5mXs/6qmf0mHj8S+DnwmZndFLf9ihCE9gSeM7NVwEWJ49+TdDAhuN2Ze7IWXk/W8JxzOFdxKjaYATsSAtE/sxvMbKGkaXF1X0Jz8C1pg1ZhLTA3sb40G8iiWUBPAEn1QG/CLOTZc6yOQSib6R7AZsAjOeepIcyUnjQpuWJmy4DpTV1ojtcSx6+RNJcwu3rW7PjaM7shBuDvA18ANo9le6VA/i25nqyRhFpyVndCbdC5ilHJwSz7Scud2DO7vROwBugXX5MWJ37OnXLOyOkTa0K2Kf4NYGbOvhU560s2KKjUH3igifxHmNmIxHq+8q7bZmYWg1CneI7jCM3EcwlBeRHwU+DAAudryfVkz7kiuS8nCDpXESo5mL1D+BAfAMyAdZPP7kJoCr0MdAZ6mtlTpZwg1vRmAV8Gnozn6EIIkC/FZJMJH+Q+ZpavCVZMKc3MluoPPGNmN2Q3xCZ4Ia25HucqVsUGMzNbJOlW4HJJ84A5wCXA2rDb3pJ0GzBW0rmE4NYDOBx43czub+aprgF+LultYAqhP2nLnHJcAVwtqRPwNJABDgYWm9mtRa6hlGZmS00HhkgaBLxHuNGxf/w5X5lKvh7nKlnFBrPoHGA0cB/QSLhbuB2wPO4/FfgFcCXweUJf2bNAcwMZ8djewBhCoLwFuAeoT6T5JSGYDif05S0g1NySzcNyGU2o/f2Z0CT9E3ADcFSRYyr5epwricxyu6QqV7wTNxM418xuLnd5OqrY3F84tGF8qr5oPnPcjWTq6po+wG0yjY2N1NfXA9SbWWOxtBVdM5O0D7A74Y5mPesfqfhr2Qrl1mkY25CqIYBqUjI2W0dVDb+984DdgJXAi0B/M/usvEVyAFPHXkMmkyl3MdqED85Y/So6mJnZy4Q7i64CZerqvFnmKkalf53JOeeaxYOZcy4VPJg551KhovvMXGVrbCx6p9y5VmvJ31hVPWfmKoOk7SnwDQPn2sm2Zpb7XeINeM3MlSL7fdJtCV9sd+WRHb0k7b+H7sShv4rxYOZaY1FTT2W79pMYvSTtv4dmXZvfAHDOpYIHM+dcKngwc6VYQRiOKe9gjm6T8d9Dgt/NdM6lgtfMnHOp4MHMOZcKHsycc6ngwcw5lwoezJxzqeDBzCHpTEnvSVou6cU432ex9ANiuuWS3o2TEOemOVbSZEkr4uu32u8K0qOtfxeSTpFkeZbN2vdKNj0PZh2cpOOBUcBvgX2Ap4AHJPUpkH4HwuxXT8X0I4BrJR2bSHMQYbaoBmCv+HqnpEITEzva53cRNRJmIFu3mNlyUsafM+vgJD0PvGRmZyS2TQHuNbPhedJfChxtZn0T20YDe5nZQXH9z0DGzI5KpHkQmG9mJ7Tf1VS3dvpdnAKMMrMt27n4Zec1sw5MUg1hjoWHc3Y9TJgUOJ+D8qR/CNhPUtcm0hTKs8Nrx98FwBaSPpD0kaT74qxnqePBrGPrAXQGZudsnw30KnBMrwLpu8T8iqUplKdrv9/FVOAU4GjgBMIE2hMl7dL6IlcWHwLIQZgJPUl5tjWVPnd7S/N0QZv+LszsOeC5dTuliYTZ638CnNWqklYYr5l1bJ8Ba9j4P39PNv6Pn/VJgfSrgblNpCmUp2u/38UGzGwt8AKQupqZB7MOzMyyEysfkbPrCOCZAoc9myf914FJZraqiTSF8uzw2vF3sQGFER33BmaVXNhKZWa+dOAFOJ4wW/xpQF/gamAx8IW4fyQwNpF+B2AJcFVMf1o8/thEmoMJtYOfAbvH11XAgeW+3kpe2ul3cREwCNiREMRuib+LA8p9vW3+/pW7AL6UfwHOBN4njIv1IvCVxL4xwISc9AMI/S4rCBObDM2T53cInc8rgSnAt8t9ndWwtPXvIgbED+L+OYS7nQeV+zrbY/HnzJxzqeB9Zs65VPBg5pxLBQ9mzrlU8GDmnEsFD2bOuVTwYOacSwUPZs65VPBg5pxLBQ9mzrlU8GDmnEsFD2bOuVT4P1MyNszYhIYFAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 200x200 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Compute Shapley Values for a given itemset\n",
    "itemset_1 = pr.iloc[1].itemsets\n",
    "itemset_shap = fp_divergence_difference.computeShapleyValue(itemset_1)\n",
    "fp_divergence_difference.plotShapleyValue(shapley_values=abbreviateDict(itemset_shap, abbreviations), \n",
    "                                          sizeFig=(2,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b4b34610",
   "metadata": {},
   "source": [
    "## Gain < 0 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "4ac034d4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_w2b</th>\n",
       "      <th>accuracy_hb</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>(action=decrease, location=washroom, ageRange=22-40)</td>\n",
       "      <td>-1.68</td>\n",
       "      <td>100.00</td>\n",
       "      <td>98.32</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>(action=decrease, gender=female, total_duration=high, total_silence=high)</td>\n",
       "      <td>-0.88</td>\n",
       "      <td>98.25</td>\n",
       "      <td>97.37</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>(trimmed_duration=high, gender=male, speed_rate_word=low, action=decrease)</td>\n",
       "      <td>-0.86</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.14</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>(trimmed_duration=high, location=washroom, object=heat, total_silence=medium)</td>\n",
       "      <td>-0.85</td>\n",
       "      <td>98.29</td>\n",
       "      <td>97.44</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>(gender=male, trimmed_duration=high, n_words=high, speed_rate_word=low, location=none)</td>\n",
       "      <td>-0.85</td>\n",
       "      <td>99.15</td>\n",
       "      <td>98.29</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>(location=washroom, n_words=high, action=increase, speed_rate_word=low)</td>\n",
       "      <td>-0.85</td>\n",
       "      <td>99.15</td>\n",
       "      <td>98.30</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>(action=decrease, object=volume, ageRange=41-65)</td>\n",
       "      <td>-0.84</td>\n",
       "      <td>97.48</td>\n",
       "      <td>96.64</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>(location=none, n_words=high, speed_rate_word=low, gender=male)</td>\n",
       "      <td>-0.83</td>\n",
       "      <td>99.17</td>\n",
       "      <td>98.33</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>(n_words=high, ageRange=41-65, speed_rate_word=low, gender=male)</td>\n",
       "      <td>-0.83</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.17</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>(gender=male, object=heat, action=decrease, n_words=high, speed_rate_word=low)</td>\n",
       "      <td>-0.83</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.17</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>(gender=male, trimmed_duration=high, object=heat, total_silence=low, ageRange=22-40, n_words=high)</td>\n",
       "      <td>-0.81</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.19</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>(trimmed_duration=high, location=washroom, speed_rate_word=low, gender=male)</td>\n",
       "      <td>-0.81</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.19</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>(trimmed_duration=high, location=none, speed_rate_word=low, gender=male)</td>\n",
       "      <td>-0.81</td>\n",
       "      <td>99.19</td>\n",
       "      <td>98.39</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>(gender=male, trimmed_duration=high, object=heat, total_silence=low, n_words=high)</td>\n",
       "      <td>-0.81</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.19</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>(gender=male, trimmed_duration=high, n_words=high, speed_rate_word=low, action=increase)</td>\n",
       "      <td>-0.80</td>\n",
       "      <td>97.60</td>\n",
       "      <td>96.80</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>(location=washroom, action=increase, speed_rate_word=low)</td>\n",
       "      <td>-0.80</td>\n",
       "      <td>98.40</td>\n",
       "      <td>97.60</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>(ageRange=41-65, speed_rate_word_trimmed=low, speed_rate_word=low, gender=male)</td>\n",
       "      <td>-0.79</td>\n",
       "      <td>99.21</td>\n",
       "      <td>98.41</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>(trimmed_duration=high, n_words=high, location=none, gender=male)</td>\n",
       "      <td>-0.77</td>\n",
       "      <td>99.23</td>\n",
       "      <td>98.46</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>(trimmed_duration=high, action=increase, speed_rate_word=low, gender=male)</td>\n",
       "      <td>-0.77</td>\n",
       "      <td>97.69</td>\n",
       "      <td>96.92</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>(location=washroom, n_words=high, speed_rate_word=low, gender=male)</td>\n",
       "      <td>-0.77</td>\n",
       "      <td>99.23</td>\n",
       "      <td>98.46</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                              itemsets  \\\n",
       "0                                                 (action=decrease, location=washroom, ageRange=22-40)   \n",
       "1                            (action=decrease, gender=female, total_duration=high, total_silence=high)   \n",
       "2                           (trimmed_duration=high, gender=male, speed_rate_word=low, action=decrease)   \n",
       "3                        (trimmed_duration=high, location=washroom, object=heat, total_silence=medium)   \n",
       "4               (gender=male, trimmed_duration=high, n_words=high, speed_rate_word=low, location=none)   \n",
       "5                              (location=washroom, n_words=high, action=increase, speed_rate_word=low)   \n",
       "6                                                     (action=decrease, object=volume, ageRange=41-65)   \n",
       "7                                      (location=none, n_words=high, speed_rate_word=low, gender=male)   \n",
       "8                                     (n_words=high, ageRange=41-65, speed_rate_word=low, gender=male)   \n",
       "9                       (gender=male, object=heat, action=decrease, n_words=high, speed_rate_word=low)   \n",
       "10  (gender=male, trimmed_duration=high, object=heat, total_silence=low, ageRange=22-40, n_words=high)   \n",
       "11                        (trimmed_duration=high, location=washroom, speed_rate_word=low, gender=male)   \n",
       "12                            (trimmed_duration=high, location=none, speed_rate_word=low, gender=male)   \n",
       "13                  (gender=male, trimmed_duration=high, object=heat, total_silence=low, n_words=high)   \n",
       "14            (gender=male, trimmed_duration=high, n_words=high, speed_rate_word=low, action=increase)   \n",
       "15                                           (location=washroom, action=increase, speed_rate_word=low)   \n",
       "16                     (ageRange=41-65, speed_rate_word_trimmed=low, speed_rate_word=low, gender=male)   \n",
       "17                                   (trimmed_duration=high, n_words=high, location=none, gender=male)   \n",
       "18                          (trimmed_duration=high, action=increase, speed_rate_word=low, gender=male)   \n",
       "19                                 (location=washroom, n_words=high, speed_rate_word=low, gender=male)   \n",
       "\n",
       "    d_difference  accuracy_w2b  accuracy_hb  support  \n",
       "0          -1.68        100.00        98.32     0.03  \n",
       "1          -0.88         98.25        97.37     0.03  \n",
       "2          -0.86        100.00        99.14     0.03  \n",
       "3          -0.85         98.29        97.44     0.03  \n",
       "4          -0.85         99.15        98.29     0.03  \n",
       "5          -0.85         99.15        98.30     0.03  \n",
       "6          -0.84         97.48        96.64     0.03  \n",
       "7          -0.83         99.17        98.33     0.03  \n",
       "8          -0.83        100.00        99.17     0.03  \n",
       "9          -0.83        100.00        99.17     0.03  \n",
       "10         -0.81        100.00        99.19     0.03  \n",
       "11         -0.81        100.00        99.19     0.03  \n",
       "12         -0.81         99.19        98.39     0.03  \n",
       "13         -0.81        100.00        99.19     0.03  \n",
       "14         -0.80         97.60        96.80     0.03  \n",
       "15         -0.80         98.40        97.60     0.03  \n",
       "16         -0.79         99.21        98.41     0.03  \n",
       "17         -0.77         99.23        98.46     0.03  \n",
       "18         -0.77         97.69        96.92     0.03  \n",
       "19         -0.77         99.23        98.46     0.03  "
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Retrieve the data cohorts for which Hubert-Base performs worse than Wav2Vec2-Base\n",
    "pr = compare_performance[cols][::-1].head(20).reset_index()\n",
    "pr[\"support\"] = pr[\"support\"].round(2)\n",
    "pr[\"accuracy_hb\"] = (pr[\"accuracy_hb\"]*100).round(2)\n",
    "pr[\"accuracy_w2b\"] = (pr[\"accuracy_w2b\"]*100).round(2)\n",
    "pr[\"d_difference\"] = (pr[\"d_difference\"]*100).round(2)\n",
    "\n",
    "pr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "7bd297bd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_w2b</th>\n",
       "      <th>accuracy_hb</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>action=decrease, ageRange=22-40, loc=washroom</td>\n",
       "      <td>-1.68</td>\n",
       "      <td>100.00</td>\n",
       "      <td>98.32</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>action=decrease, gender=female, dur=high, tot_silence=high</td>\n",
       "      <td>-0.88</td>\n",
       "      <td>98.25</td>\n",
       "      <td>97.37</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     itemsets  d_difference  \\\n",
       "0               action=decrease, ageRange=22-40, loc=washroom         -1.68   \n",
       "1  action=decrease, gender=female, dur=high, tot_silence=high         -0.88   \n",
       "\n",
       "   accuracy_w2b  accuracy_hb  support  \n",
       "0        100.00        98.32     0.03  \n",
       "1         98.25        97.37     0.03  "
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Abbreviate itemset names for better visualization\n",
    "pr_l = pr.head(2).copy()\n",
    "pr_l['itemsets'] = pr_l['itemsets'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "pr_l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "96b57661",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 200x200 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Compute Shapley Values for a given itemset\n",
    "itemset_1 = pr.iloc[0].itemsets\n",
    "itemset_shap = fp_divergence_difference.computeShapleyValue(itemset_1)\n",
    "fp_divergence_difference.plotShapleyValue(shapley_values=abbreviateDict(itemset_shap, abbreviations), \n",
    "                                          sizeFig=(2,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a566eb0",
   "metadata": {},
   "source": [
    "## Gain = 0 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "d6fb950a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_w2b</th>\n",
       "      <th>accuracy_hb</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4785</th>\n",
       "      <td>(action=decrease, n_words=high, speed_rate_word_trimmed=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>95.49</td>\n",
       "      <td>95.49</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4784</th>\n",
       "      <td>(action=decrease, trimmed_duration=high, speed_rate_word_trimmed=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>96.23</td>\n",
       "      <td>96.23</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4783</th>\n",
       "      <td>(action=decrease, n_words=high, total_duration=high, gender=female)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>96.43</td>\n",
       "      <td>96.43</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4782</th>\n",
       "      <td>(action=decrease, gender=male, object=heat, total_silence=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>96.97</td>\n",
       "      <td>96.97</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4781</th>\n",
       "      <td>(action=decrease, gender=female, trimmed_duration=high, total_duration=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.13</td>\n",
       "      <td>97.13</td>\n",
       "      <td>0.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4780</th>\n",
       "      <td>(trimmed_duration=high, action=increase, speed_rate_word_trimmed=low, speed_rate_word=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.18</td>\n",
       "      <td>97.18</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4779</th>\n",
       "      <td>(action=decrease, ageRange=22-40, object=heat, total_silence=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.30</td>\n",
       "      <td>97.30</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4778</th>\n",
       "      <td>(ageRange=22-40, object=heat, speed_rate_word_trimmed=medium, total_silence=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.44</td>\n",
       "      <td>97.44</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4777</th>\n",
       "      <td>(action=decrease, n_words=high, ageRange=22-40, total_duration=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.50</td>\n",
       "      <td>97.50</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4776</th>\n",
       "      <td>(trimmed_duration=high, ageRange=22-40, action=increase, speed_rate_word=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.50</td>\n",
       "      <td>97.50</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4775</th>\n",
       "      <td>(gender=male, trimmed_duration=high, object=heat, speed_rate_word_trimmed=low, speed_rate_word=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.52</td>\n",
       "      <td>97.52</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4774</th>\n",
       "      <td>(location=washroom, trimmed_duration=medium, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.54</td>\n",
       "      <td>97.54</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4773</th>\n",
       "      <td>(location=washroom, ageRange=22-40, trimmed_duration=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.54</td>\n",
       "      <td>97.54</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4772</th>\n",
       "      <td>(n_words=high, object=heat, total_duration=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.60</td>\n",
       "      <td>97.60</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4771</th>\n",
       "      <td>(ageRange=22-40, n_words=high, total_duration=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.73</td>\n",
       "      <td>97.73</td>\n",
       "      <td>0.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4770</th>\n",
       "      <td>(action=decrease, gender=female, ageRange=22-40, total_duration=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.78</td>\n",
       "      <td>97.78</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4769</th>\n",
       "      <td>(ageRange=22-40, speed_rate_word_trimmed=medium, object=heat, speed_rate_word=low)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.84</td>\n",
       "      <td>97.84</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4768</th>\n",
       "      <td>(location=kitchen, n_words=medium)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.84</td>\n",
       "      <td>97.84</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4767</th>\n",
       "      <td>(n_words=high, total_duration=medium, gender=male)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.86</td>\n",
       "      <td>97.86</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4766</th>\n",
       "      <td>(location=washroom, speed_rate_word_trimmed=high)</td>\n",
       "      <td>0.0</td>\n",
       "      <td>97.90</td>\n",
       "      <td>97.90</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                 itemsets  \\\n",
       "4785                                      (action=decrease, n_words=high, speed_rate_word_trimmed=medium)   \n",
       "4784                             (action=decrease, trimmed_duration=high, speed_rate_word_trimmed=medium)   \n",
       "4783                                  (action=decrease, n_words=high, total_duration=high, gender=female)   \n",
       "4782                                       (action=decrease, gender=male, object=heat, total_silence=low)   \n",
       "4781                         (action=decrease, gender=female, trimmed_duration=high, total_duration=high)   \n",
       "4780           (trimmed_duration=high, action=increase, speed_rate_word_trimmed=low, speed_rate_word=low)   \n",
       "4779                                    (action=decrease, ageRange=22-40, object=heat, total_silence=low)   \n",
       "4778                     (ageRange=22-40, object=heat, speed_rate_word_trimmed=medium, total_silence=low)   \n",
       "4777                                 (action=decrease, n_words=high, ageRange=22-40, total_duration=high)   \n",
       "4776                        (trimmed_duration=high, ageRange=22-40, action=increase, speed_rate_word=low)   \n",
       "4775  (gender=male, trimmed_duration=high, object=heat, speed_rate_word_trimmed=low, speed_rate_word=low)   \n",
       "4774                                            (location=washroom, trimmed_duration=medium, gender=male)   \n",
       "4773                                         (location=washroom, ageRange=22-40, trimmed_duration=medium)   \n",
       "4772                                                   (n_words=high, object=heat, total_duration=medium)   \n",
       "4771                                                (ageRange=22-40, n_words=high, total_duration=medium)   \n",
       "4770                                (action=decrease, gender=female, ageRange=22-40, total_duration=high)   \n",
       "4769                   (ageRange=22-40, speed_rate_word_trimmed=medium, object=heat, speed_rate_word=low)   \n",
       "4768                                                                   (location=kitchen, n_words=medium)   \n",
       "4767                                                   (n_words=high, total_duration=medium, gender=male)   \n",
       "4766                                                    (location=washroom, speed_rate_word_trimmed=high)   \n",
       "\n",
       "      d_difference  accuracy_w2b  accuracy_hb  support  \n",
       "4785           0.0         95.49        95.49     0.04  \n",
       "4784           0.0         96.23        96.23     0.04  \n",
       "4783           0.0         96.43        96.43     0.04  \n",
       "4782           0.0         96.97        96.97     0.03  \n",
       "4781           0.0         97.13        97.13     0.05  \n",
       "4780           0.0         97.18        97.18     0.04  \n",
       "4779           0.0         97.30        97.30     0.04  \n",
       "4778           0.0         97.44        97.44     0.03  \n",
       "4777           0.0         97.50        97.50     0.03  \n",
       "4776           0.0         97.50        97.50     0.04  \n",
       "4775           0.0         97.52        97.52     0.03  \n",
       "4774           0.0         97.54        97.54     0.03  \n",
       "4773           0.0         97.54        97.54     0.03  \n",
       "4772           0.0         97.60        97.60     0.03  \n",
       "4771           0.0         97.73        97.73     0.03  \n",
       "4770           0.0         97.78        97.78     0.04  \n",
       "4769           0.0         97.84        97.84     0.04  \n",
       "4768           0.0         97.84        97.84     0.04  \n",
       "4767           0.0         97.86        97.86     0.04  \n",
       "4766           0.0         97.90        97.90     0.04  "
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Retrieve the data cohorts for which Hubert-Base performs equal to Wav2Vec2-Base\n",
    "pr = merged.loc[ fp_divergence_difference.getDivergence(th_redundancy=0.0).itemsets.values][cols].reset_index()\n",
    "pr[\"support\"] = pr[\"support\"].round(2)\n",
    "pr[\"accuracy_hb\"] = (pr[\"accuracy_hb\"]*100).round(2)\n",
    "pr[\"accuracy_w2b\"] = (pr[\"accuracy_w2b\"]*100).round(2)\n",
    "pr[\"d_difference\"] = (pr[\"d_difference\"]*100).round(2)\n",
    "pr = pr.loc[abs(pr[\"d_difference\"])==0]\n",
    "\n",
    "pr = pr.sort_values(\"d_difference\").sort_values(\"accuracy_hb\")\n",
    "pr.head(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "77fef497",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>itemsets</th>\n",
       "      <th>d_difference</th>\n",
       "      <th>accuracy_w2b</th>\n",
       "      <th>accuracy_hb</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4785</th>\n",
       "      <td>action=decrease, n_words=high, speakRate_trim=medium</td>\n",
       "      <td>0.0</td>\n",
       "      <td>95.49</td>\n",
       "      <td>95.49</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4784</th>\n",
       "      <td>action=decrease, speakRate_trim=medium, trim_dur=high</td>\n",
       "      <td>0.0</td>\n",
       "      <td>96.23</td>\n",
       "      <td>96.23</td>\n",
       "      <td>0.04</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                   itemsets  d_difference  \\\n",
       "4785   action=decrease, n_words=high, speakRate_trim=medium           0.0   \n",
       "4784  action=decrease, speakRate_trim=medium, trim_dur=high           0.0   \n",
       "\n",
       "      accuracy_w2b  accuracy_hb  support  \n",
       "4785         95.49        95.49     0.04  \n",
       "4784         96.23        96.23     0.04  "
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Abbreviate itemset names for better visualization\n",
    "pr_l = pr.head(2).copy()\n",
    "pr_l['itemsets'] = pr_l['itemsets'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "pr_l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "d17cfdc7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 200x200 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Compute Shapley Values for a given itemset\n",
    "itemset_1 = pr.iloc[0].itemsets\n",
    "itemset_shap = fp_divergence_difference.computeShapleyValue(itemset_1)\n",
    "fp_divergence_difference.plotShapleyValue(shapley_values=abbreviateDict(itemset_shap, abbreviations), \n",
    "                                          sizeFig=(2,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9707900a",
   "metadata": {},
   "source": [
    "## Inspect gender"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b969a5e2",
   "metadata": {},
   "source": [
    "We consider subgroups among the top-10 of each gain category for which both two genders are above support."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "3464ced3",
   "metadata": {},
   "outputs": [],
   "source": [
    "n = 40\n",
    "cols = ['d_difference','accuracy_w2b', 'accuracy_hb', 'support']\n",
    "comparison_res = []\n",
    "config1 = \"wav2vec2-base\"\n",
    "config2 = \"hubert-base\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "56047630",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Gain</th>\n",
       "      <th>Subgroup</th>\n",
       "      <th>wav2vec2-base_gender=female</th>\n",
       "      <th>hubert-base_gender=female</th>\n",
       "      <th>sup_gender=female</th>\n",
       "      <th>wav2vec2-base_gender=male</th>\n",
       "      <th>hubert-base_gender=male</th>\n",
       "      <th>sup_gender=male</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(n_words=low, location=none, total_silence=high, trimmed_duration=low)</td>\n",
       "      <td>80.70</td>\n",
       "      <td>97.66</td>\n",
       "      <td>171.0</td>\n",
       "      <td>64.00</td>\n",
       "      <td>95.20</td>\n",
       "      <td>125.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(n_words=low, total_silence=high, location=none, speed_rate_word=high, trimmed_duration=low)</td>\n",
       "      <td>80.70</td>\n",
       "      <td>97.66</td>\n",
       "      <td>171.0</td>\n",
       "      <td>64.23</td>\n",
       "      <td>95.12</td>\n",
       "      <td>123.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(ageRange=22-40, object=volume, speed_rate_word=high)</td>\n",
       "      <td>83.33</td>\n",
       "      <td>96.80</td>\n",
       "      <td>156.0</td>\n",
       "      <td>67.52</td>\n",
       "      <td>98.29</td>\n",
       "      <td>117.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(n_words=low, trimmed_duration=low, total_silence=high)</td>\n",
       "      <td>83.33</td>\n",
       "      <td>97.98</td>\n",
       "      <td>198.0</td>\n",
       "      <td>65.04</td>\n",
       "      <td>95.80</td>\n",
       "      <td>143.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(total_silence=high, n_words=low, speed_rate_word=high, trimmed_duration=low)</td>\n",
       "      <td>83.33</td>\n",
       "      <td>97.98</td>\n",
       "      <td>198.0</td>\n",
       "      <td>65.25</td>\n",
       "      <td>95.74</td>\n",
       "      <td>141.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(ageRange=22-40, total_silence=high, location=none, speed_rate_word=high)</td>\n",
       "      <td>91.07</td>\n",
       "      <td>98.21</td>\n",
       "      <td>168.0</td>\n",
       "      <td>60.50</td>\n",
       "      <td>89.92</td>\n",
       "      <td>119.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(total_silence=high, location=none, speed_rate_word=high, trimmed_duration=low)</td>\n",
       "      <td>82.54</td>\n",
       "      <td>97.88</td>\n",
       "      <td>189.0</td>\n",
       "      <td>64.66</td>\n",
       "      <td>93.98</td>\n",
       "      <td>133.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(total_silence=high, trimmed_duration=low, speed_rate_word=high)</td>\n",
       "      <td>84.72</td>\n",
       "      <td>98.15</td>\n",
       "      <td>216.0</td>\n",
       "      <td>65.56</td>\n",
       "      <td>94.70</td>\n",
       "      <td>151.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(ageRange=22-40, n_words=low, location=none, speed_rate_word=high, trimmed_duration=low)</td>\n",
       "      <td>84.18</td>\n",
       "      <td>96.04</td>\n",
       "      <td>177.0</td>\n",
       "      <td>67.63</td>\n",
       "      <td>96.62</td>\n",
       "      <td>207.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(ageRange=22-40, n_words=low, speed_rate_word=high, trimmed_duration=low)</td>\n",
       "      <td>85.86</td>\n",
       "      <td>96.46</td>\n",
       "      <td>198.0</td>\n",
       "      <td>68.16</td>\n",
       "      <td>96.86</td>\n",
       "      <td>223.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(ageRange=22-40, total_silence=high, speed_rate_word=high)</td>\n",
       "      <td>92.50</td>\n",
       "      <td>98.50</td>\n",
       "      <td>200.0</td>\n",
       "      <td>62.86</td>\n",
       "      <td>91.43</td>\n",
       "      <td>140.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(ageRange=22-40, location=none, speed_rate_word=high, trimmed_duration=low)</td>\n",
       "      <td>85.03</td>\n",
       "      <td>96.26</td>\n",
       "      <td>187.0</td>\n",
       "      <td>67.44</td>\n",
       "      <td>95.81</td>\n",
       "      <td>215.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(ageRange=22-40, trimmed_duration=low, speed_rate_word=high)</td>\n",
       "      <td>86.54</td>\n",
       "      <td>96.64</td>\n",
       "      <td>208.0</td>\n",
       "      <td>67.96</td>\n",
       "      <td>96.10</td>\n",
       "      <td>231.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(n_words=low, location=none, total_silence=high)</td>\n",
       "      <td>85.26</td>\n",
       "      <td>98.41</td>\n",
       "      <td>251.0</td>\n",
       "      <td>67.81</td>\n",
       "      <td>95.89</td>\n",
       "      <td>146.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(n_words=low, total_silence=high, location=none, speed_rate_word=high)</td>\n",
       "      <td>85.26</td>\n",
       "      <td>98.41</td>\n",
       "      <td>251.0</td>\n",
       "      <td>68.06</td>\n",
       "      <td>95.83</td>\n",
       "      <td>144.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>(n_words=low, location=none, action=increase, trimmed_duration=low)</td>\n",
       "      <td>68.29</td>\n",
       "      <td>95.12</td>\n",
       "      <td>123.0</td>\n",
       "      <td>73.48</td>\n",
       "      <td>98.48</td>\n",
       "      <td>132.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>=</td>\n",
       "      <td>(location=none, trimmed_duration=high)</td>\n",
       "      <td>96.17</td>\n",
       "      <td>99.62</td>\n",
       "      <td>261.0</td>\n",
       "      <td>98.82</td>\n",
       "      <td>98.82</td>\n",
       "      <td>170.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>=</td>\n",
       "      <td>(n_words=high, speed_rate_word_trimmed=medium, speed_rate_word=low)</td>\n",
       "      <td>93.18</td>\n",
       "      <td>97.73</td>\n",
       "      <td>132.0</td>\n",
       "      <td>99.27</td>\n",
       "      <td>99.27</td>\n",
       "      <td>137.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(n_words=high, ageRange=41-65, speed_rate_word=low)</td>\n",
       "      <td>93.84</td>\n",
       "      <td>98.63</td>\n",
       "      <td>146.0</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.17</td>\n",
       "      <td>120.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(ageRange=41-65, speed_rate_word_trimmed=low, speed_rate_word=low)</td>\n",
       "      <td>91.52</td>\n",
       "      <td>98.18</td>\n",
       "      <td>165.0</td>\n",
       "      <td>99.21</td>\n",
       "      <td>98.41</td>\n",
       "      <td>126.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(trimmed_duration=high, n_words=high, location=none)</td>\n",
       "      <td>96.48</td>\n",
       "      <td>99.30</td>\n",
       "      <td>142.0</td>\n",
       "      <td>99.23</td>\n",
       "      <td>98.46</td>\n",
       "      <td>130.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(n_words=high, ageRange=41-65)</td>\n",
       "      <td>93.27</td>\n",
       "      <td>98.66</td>\n",
       "      <td>223.0</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.24</td>\n",
       "      <td>131.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(location=none, n_words=high)</td>\n",
       "      <td>96.48</td>\n",
       "      <td>99.30</td>\n",
       "      <td>142.0</td>\n",
       "      <td>99.25</td>\n",
       "      <td>98.51</td>\n",
       "      <td>134.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>(ageRange=41-65, speed_rate_word=low)</td>\n",
       "      <td>91.70</td>\n",
       "      <td>97.92</td>\n",
       "      <td>289.0</td>\n",
       "      <td>99.32</td>\n",
       "      <td>98.65</td>\n",
       "      <td>148.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Gain  \\\n",
       "0     >   \n",
       "1     >   \n",
       "2     >   \n",
       "3     >   \n",
       "4     >   \n",
       "5     >   \n",
       "6     >   \n",
       "7     >   \n",
       "8     >   \n",
       "9     >   \n",
       "10    >   \n",
       "11    >   \n",
       "12    >   \n",
       "13    >   \n",
       "14    >   \n",
       "15    >   \n",
       "0     =   \n",
       "1     =   \n",
       "0     <   \n",
       "1     <   \n",
       "2     <   \n",
       "3     <   \n",
       "4     <   \n",
       "5     <   \n",
       "\n",
       "                                                                                        Subgroup  \\\n",
       "0                         (n_words=low, location=none, total_silence=high, trimmed_duration=low)   \n",
       "1   (n_words=low, total_silence=high, location=none, speed_rate_word=high, trimmed_duration=low)   \n",
       "2                                          (ageRange=22-40, object=volume, speed_rate_word=high)   \n",
       "3                                        (n_words=low, trimmed_duration=low, total_silence=high)   \n",
       "4                  (total_silence=high, n_words=low, speed_rate_word=high, trimmed_duration=low)   \n",
       "5                      (ageRange=22-40, total_silence=high, location=none, speed_rate_word=high)   \n",
       "6                (total_silence=high, location=none, speed_rate_word=high, trimmed_duration=low)   \n",
       "7                               (total_silence=high, trimmed_duration=low, speed_rate_word=high)   \n",
       "8       (ageRange=22-40, n_words=low, location=none, speed_rate_word=high, trimmed_duration=low)   \n",
       "9                      (ageRange=22-40, n_words=low, speed_rate_word=high, trimmed_duration=low)   \n",
       "10                                    (ageRange=22-40, total_silence=high, speed_rate_word=high)   \n",
       "11                   (ageRange=22-40, location=none, speed_rate_word=high, trimmed_duration=low)   \n",
       "12                                  (ageRange=22-40, trimmed_duration=low, speed_rate_word=high)   \n",
       "13                                              (n_words=low, location=none, total_silence=high)   \n",
       "14                        (n_words=low, total_silence=high, location=none, speed_rate_word=high)   \n",
       "15                           (n_words=low, location=none, action=increase, trimmed_duration=low)   \n",
       "0                                                         (location=none, trimmed_duration=high)   \n",
       "1                            (n_words=high, speed_rate_word_trimmed=medium, speed_rate_word=low)   \n",
       "0                                            (n_words=high, ageRange=41-65, speed_rate_word=low)   \n",
       "1                             (ageRange=41-65, speed_rate_word_trimmed=low, speed_rate_word=low)   \n",
       "2                                           (trimmed_duration=high, n_words=high, location=none)   \n",
       "3                                                                 (n_words=high, ageRange=41-65)   \n",
       "4                                                                  (location=none, n_words=high)   \n",
       "5                                                          (ageRange=41-65, speed_rate_word=low)   \n",
       "\n",
       "    wav2vec2-base_gender=female  hubert-base_gender=female  sup_gender=female  \\\n",
       "0                         80.70                      97.66              171.0   \n",
       "1                         80.70                      97.66              171.0   \n",
       "2                         83.33                      96.80              156.0   \n",
       "3                         83.33                      97.98              198.0   \n",
       "4                         83.33                      97.98              198.0   \n",
       "5                         91.07                      98.21              168.0   \n",
       "6                         82.54                      97.88              189.0   \n",
       "7                         84.72                      98.15              216.0   \n",
       "8                         84.18                      96.04              177.0   \n",
       "9                         85.86                      96.46              198.0   \n",
       "10                        92.50                      98.50              200.0   \n",
       "11                        85.03                      96.26              187.0   \n",
       "12                        86.54                      96.64              208.0   \n",
       "13                        85.26                      98.41              251.0   \n",
       "14                        85.26                      98.41              251.0   \n",
       "15                        68.29                      95.12              123.0   \n",
       "0                         96.17                      99.62              261.0   \n",
       "1                         93.18                      97.73              132.0   \n",
       "0                         93.84                      98.63              146.0   \n",
       "1                         91.52                      98.18              165.0   \n",
       "2                         96.48                      99.30              142.0   \n",
       "3                         93.27                      98.66              223.0   \n",
       "4                         96.48                      99.30              142.0   \n",
       "5                         91.70                      97.92              289.0   \n",
       "\n",
       "    wav2vec2-base_gender=male  hubert-base_gender=male  sup_gender=male  \n",
       "0                       64.00                    95.20            125.0  \n",
       "1                       64.23                    95.12            123.0  \n",
       "2                       67.52                    98.29            117.0  \n",
       "3                       65.04                    95.80            143.0  \n",
       "4                       65.25                    95.74            141.0  \n",
       "5                       60.50                    89.92            119.0  \n",
       "6                       64.66                    93.98            133.0  \n",
       "7                       65.56                    94.70            151.0  \n",
       "8                       67.63                    96.62            207.0  \n",
       "9                       68.16                    96.86            223.0  \n",
       "10                      62.86                    91.43            140.0  \n",
       "11                      67.44                    95.81            215.0  \n",
       "12                      67.96                    96.10            231.0  \n",
       "13                      67.81                    95.89            146.0  \n",
       "14                      68.06                    95.83            144.0  \n",
       "15                      73.48                    98.48            132.0  \n",
       "0                       98.82                    98.82            170.0  \n",
       "1                       99.27                    99.27            137.0  \n",
       "0                      100.00                    99.17            120.0  \n",
       "1                       99.21                    98.41            126.0  \n",
       "2                       99.23                    98.46            130.0  \n",
       "3                      100.00                    99.24            131.0  \n",
       "4                       99.25                    98.51            134.0  \n",
       "5                       99.32                    98.65            148.0  "
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Compute gain in performance from Wav2Vec2-Base to Hubert-Base, when considering different genders\n",
    "great = compare_performance[cols].head(n).reset_index()\n",
    "notr = merged.loc[ fp_divergence_difference.getDivergence(th_redundancy=0.0).itemsets.values][cols].reset_index()\n",
    "equal = notr.loc[notr[diff]==0].sort_values('accuracy_hb').head(n)\n",
    "low = compare_performance[cols][::-1].head(n).reset_index()\n",
    "\n",
    "for gain_type, pr in zip(['>', \"=\", \"<\"], [great, equal, low]):\n",
    "    gain_df_gender = filter_itemset_df_by_attributes(pr, ['gender'], alls=False)\n",
    "\n",
    "    itemsets = gain_df_gender.itemsets.values\n",
    "\n",
    "    get_accuracy = lambda dfi: sum(dfi['prediction'].values.astype(int))/len(dfi)\n",
    "    get_support_count = lambda dfi: len(dfi)\n",
    "\n",
    "    res2 = []\n",
    "    res = []\n",
    "\n",
    "    genders = ['gender=female', 'gender=male']\n",
    "    configs = [config1, config2]\n",
    "    for itemset_i in itemsets:\n",
    "        itemset_i = itemset_i.difference(genders)\n",
    "        r2 = [gain_type, itemset_i]\n",
    "        both_genders = []\n",
    "        both_genders_2 = []\n",
    "        for gender in genders:\n",
    "            itemset = itemset_i.union([gender])\n",
    "            r = [gain_type, itemset]\n",
    "\n",
    "            for config in configs:\n",
    "                if itemset in  fp_divergence_dict[config].itemset_divergence[len(itemset)]:\n",
    "                    info = fp_divergence_dict[config].getInfoItemset(itemset).iloc[0]\n",
    "                    acc = info[\"accuracy\"]\n",
    "                    s = info[\"support_count\"]\n",
    "\n",
    "                    r.append((acc*100).round(2))\n",
    "                    r2.append((acc*100).round(2))\n",
    "                    r.append(s)\n",
    "\n",
    "\n",
    "            if len(r)>2:\n",
    "                r2.append(s)\n",
    "                both_genders.append(r)\n",
    "                both_genders_2.append(r2)\n",
    "\n",
    "        if len(both_genders)==2:\n",
    "            for r in both_genders:\n",
    "                res.append(r)\n",
    "            res2.append(r2)\n",
    "    \n",
    "    cols =[]\n",
    "    for gender in genders:\n",
    "        cols.extend([f\"{c}_{gender}\" for c in configs + [\"sup\"]])\n",
    "    v = pd.DataFrame(res, columns = ['Gain',\"Subgroup\"] + configs + [\"S1\", \"S2\"])\n",
    "    v2 = pd.DataFrame(res2, columns = ['Gain', \"Subgroup\"] + cols )\n",
    "\n",
    "    comparison_res.append(v2)\n",
    "\n",
    "comparison_res = pd.concat(comparison_res)\n",
    "comparison_res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "c632adb1",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Gain</th>\n",
       "      <th>Subgroup_str</th>\n",
       "      <th>wav2vec2-base_gender=female</th>\n",
       "      <th>hubert-base_gender=female</th>\n",
       "      <th>wav2vec2-base_gender=male</th>\n",
       "      <th>hubert-base_gender=male</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>loc=none, n_words=low, tot_silence=high, trim_dur=low</td>\n",
       "      <td>80.70</td>\n",
       "      <td>97.66</td>\n",
       "      <td>64.00</td>\n",
       "      <td>95.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>loc=none, n_words=low, speakRate=high, tot_silence=high, trim_dur=low</td>\n",
       "      <td>80.70</td>\n",
       "      <td>97.66</td>\n",
       "      <td>64.23</td>\n",
       "      <td>95.12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>ageRange=22-40, object=volume, speakRate=high</td>\n",
       "      <td>83.33</td>\n",
       "      <td>96.80</td>\n",
       "      <td>67.52</td>\n",
       "      <td>98.29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>n_words=low, tot_silence=high, trim_dur=low</td>\n",
       "      <td>83.33</td>\n",
       "      <td>97.98</td>\n",
       "      <td>65.04</td>\n",
       "      <td>95.80</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>n_words=low, speakRate=high, tot_silence=high, trim_dur=low</td>\n",
       "      <td>83.33</td>\n",
       "      <td>97.98</td>\n",
       "      <td>65.25</td>\n",
       "      <td>95.74</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>ageRange=22-40, loc=none, speakRate=high, tot_silence=high</td>\n",
       "      <td>91.07</td>\n",
       "      <td>98.21</td>\n",
       "      <td>60.50</td>\n",
       "      <td>89.92</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>loc=none, speakRate=high, tot_silence=high, trim_dur=low</td>\n",
       "      <td>82.54</td>\n",
       "      <td>97.88</td>\n",
       "      <td>64.66</td>\n",
       "      <td>93.98</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>speakRate=high, tot_silence=high, trim_dur=low</td>\n",
       "      <td>84.72</td>\n",
       "      <td>98.15</td>\n",
       "      <td>65.56</td>\n",
       "      <td>94.70</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>ageRange=22-40, loc=none, n_words=low, speakRate=high, trim_dur=low</td>\n",
       "      <td>84.18</td>\n",
       "      <td>96.04</td>\n",
       "      <td>67.63</td>\n",
       "      <td>96.62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>ageRange=22-40, n_words=low, speakRate=high, trim_dur=low</td>\n",
       "      <td>85.86</td>\n",
       "      <td>96.46</td>\n",
       "      <td>68.16</td>\n",
       "      <td>96.86</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>ageRange=22-40, speakRate=high, tot_silence=high</td>\n",
       "      <td>92.50</td>\n",
       "      <td>98.50</td>\n",
       "      <td>62.86</td>\n",
       "      <td>91.43</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>ageRange=22-40, loc=none, speakRate=high, trim_dur=low</td>\n",
       "      <td>85.03</td>\n",
       "      <td>96.26</td>\n",
       "      <td>67.44</td>\n",
       "      <td>95.81</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>ageRange=22-40, speakRate=high, trim_dur=low</td>\n",
       "      <td>86.54</td>\n",
       "      <td>96.64</td>\n",
       "      <td>67.96</td>\n",
       "      <td>96.10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>loc=none, n_words=low, tot_silence=high</td>\n",
       "      <td>85.26</td>\n",
       "      <td>98.41</td>\n",
       "      <td>67.81</td>\n",
       "      <td>95.89</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>loc=none, n_words=low, speakRate=high, tot_silence=high</td>\n",
       "      <td>85.26</td>\n",
       "      <td>98.41</td>\n",
       "      <td>68.06</td>\n",
       "      <td>95.83</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>action=increase, loc=none, n_words=low, trim_dur=low</td>\n",
       "      <td>68.29</td>\n",
       "      <td>95.12</td>\n",
       "      <td>73.48</td>\n",
       "      <td>98.48</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>=</td>\n",
       "      <td>loc=none, trim_dur=high</td>\n",
       "      <td>96.17</td>\n",
       "      <td>99.62</td>\n",
       "      <td>98.82</td>\n",
       "      <td>98.82</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>=</td>\n",
       "      <td>n_words=high, speakRate=low, speakRate_trim=medium</td>\n",
       "      <td>93.18</td>\n",
       "      <td>97.73</td>\n",
       "      <td>99.27</td>\n",
       "      <td>99.27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>ageRange=41-65, n_words=high, speakRate=low</td>\n",
       "      <td>93.84</td>\n",
       "      <td>98.63</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>ageRange=41-65, speakRate=low, speakRate_trim=low</td>\n",
       "      <td>91.52</td>\n",
       "      <td>98.18</td>\n",
       "      <td>99.21</td>\n",
       "      <td>98.41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>loc=none, n_words=high, trim_dur=high</td>\n",
       "      <td>96.48</td>\n",
       "      <td>99.30</td>\n",
       "      <td>99.23</td>\n",
       "      <td>98.46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>ageRange=41-65, n_words=high</td>\n",
       "      <td>93.27</td>\n",
       "      <td>98.66</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>loc=none, n_words=high</td>\n",
       "      <td>96.48</td>\n",
       "      <td>99.30</td>\n",
       "      <td>99.25</td>\n",
       "      <td>98.51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>ageRange=41-65, speakRate=low</td>\n",
       "      <td>91.70</td>\n",
       "      <td>97.92</td>\n",
       "      <td>99.32</td>\n",
       "      <td>98.65</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Gain  \\\n",
       "0     >   \n",
       "1     >   \n",
       "2     >   \n",
       "3     >   \n",
       "4     >   \n",
       "5     >   \n",
       "6     >   \n",
       "7     >   \n",
       "8     >   \n",
       "9     >   \n",
       "10    >   \n",
       "11    >   \n",
       "12    >   \n",
       "13    >   \n",
       "14    >   \n",
       "15    >   \n",
       "0     =   \n",
       "1     =   \n",
       "0     <   \n",
       "1     <   \n",
       "2     <   \n",
       "3     <   \n",
       "4     <   \n",
       "5     <   \n",
       "\n",
       "                                                             Subgroup_str  \\\n",
       "0                   loc=none, n_words=low, tot_silence=high, trim_dur=low   \n",
       "1   loc=none, n_words=low, speakRate=high, tot_silence=high, trim_dur=low   \n",
       "2                           ageRange=22-40, object=volume, speakRate=high   \n",
       "3                             n_words=low, tot_silence=high, trim_dur=low   \n",
       "4             n_words=low, speakRate=high, tot_silence=high, trim_dur=low   \n",
       "5              ageRange=22-40, loc=none, speakRate=high, tot_silence=high   \n",
       "6                loc=none, speakRate=high, tot_silence=high, trim_dur=low   \n",
       "7                          speakRate=high, tot_silence=high, trim_dur=low   \n",
       "8     ageRange=22-40, loc=none, n_words=low, speakRate=high, trim_dur=low   \n",
       "9               ageRange=22-40, n_words=low, speakRate=high, trim_dur=low   \n",
       "10                       ageRange=22-40, speakRate=high, tot_silence=high   \n",
       "11                 ageRange=22-40, loc=none, speakRate=high, trim_dur=low   \n",
       "12                           ageRange=22-40, speakRate=high, trim_dur=low   \n",
       "13                                loc=none, n_words=low, tot_silence=high   \n",
       "14                loc=none, n_words=low, speakRate=high, tot_silence=high   \n",
       "15                   action=increase, loc=none, n_words=low, trim_dur=low   \n",
       "0                                                 loc=none, trim_dur=high   \n",
       "1                      n_words=high, speakRate=low, speakRate_trim=medium   \n",
       "0                             ageRange=41-65, n_words=high, speakRate=low   \n",
       "1                       ageRange=41-65, speakRate=low, speakRate_trim=low   \n",
       "2                                   loc=none, n_words=high, trim_dur=high   \n",
       "3                                            ageRange=41-65, n_words=high   \n",
       "4                                                  loc=none, n_words=high   \n",
       "5                                           ageRange=41-65, speakRate=low   \n",
       "\n",
       "    wav2vec2-base_gender=female  hubert-base_gender=female  \\\n",
       "0                         80.70                      97.66   \n",
       "1                         80.70                      97.66   \n",
       "2                         83.33                      96.80   \n",
       "3                         83.33                      97.98   \n",
       "4                         83.33                      97.98   \n",
       "5                         91.07                      98.21   \n",
       "6                         82.54                      97.88   \n",
       "7                         84.72                      98.15   \n",
       "8                         84.18                      96.04   \n",
       "9                         85.86                      96.46   \n",
       "10                        92.50                      98.50   \n",
       "11                        85.03                      96.26   \n",
       "12                        86.54                      96.64   \n",
       "13                        85.26                      98.41   \n",
       "14                        85.26                      98.41   \n",
       "15                        68.29                      95.12   \n",
       "0                         96.17                      99.62   \n",
       "1                         93.18                      97.73   \n",
       "0                         93.84                      98.63   \n",
       "1                         91.52                      98.18   \n",
       "2                         96.48                      99.30   \n",
       "3                         93.27                      98.66   \n",
       "4                         96.48                      99.30   \n",
       "5                         91.70                      97.92   \n",
       "\n",
       "    wav2vec2-base_gender=male  hubert-base_gender=male  \n",
       "0                       64.00                    95.20  \n",
       "1                       64.23                    95.12  \n",
       "2                       67.52                    98.29  \n",
       "3                       65.04                    95.80  \n",
       "4                       65.25                    95.74  \n",
       "5                       60.50                    89.92  \n",
       "6                       64.66                    93.98  \n",
       "7                       65.56                    94.70  \n",
       "8                       67.63                    96.62  \n",
       "9                       68.16                    96.86  \n",
       "10                      62.86                    91.43  \n",
       "11                      67.44                    95.81  \n",
       "12                      67.96                    96.10  \n",
       "13                      67.81                    95.89  \n",
       "14                      68.06                    95.83  \n",
       "15                      73.48                    98.48  \n",
       "0                       98.82                    98.82  \n",
       "1                       99.27                    99.27  \n",
       "0                      100.00                    99.17  \n",
       "1                       99.21                    98.41  \n",
       "2                       99.23                    98.46  \n",
       "3                      100.00                    99.24  \n",
       "4                       99.25                    98.51  \n",
       "5                       99.32                    98.65  "
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Select columns for better visualization\n",
    "comparison_res_pr = comparison_res.copy()\n",
    "\n",
    "comparison_res_pr['Subgroup_str'] = comparison_res_pr['Subgroup'].apply(lambda x: sortItemset(x, abbreviations))\n",
    "\n",
    "comparison_res_pr = comparison_res_pr[['Gain', 'Subgroup_str', \n",
    "       'wav2vec2-base_gender=female', 'hubert-base_gender=female',\n",
    "       'wav2vec2-base_gender=male', 'hubert-base_gender=male']]\n",
    "\n",
    "comparison_res_pr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "8639411c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Gain</th>\n",
       "      <th>Subgroup_str</th>\n",
       "      <th>wav2vec2-base_gender=female</th>\n",
       "      <th>hubert-base_gender=female</th>\n",
       "      <th>wav2vec2-base_gender=male</th>\n",
       "      <th>hubert-base_gender=male</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&gt;</td>\n",
       "      <td>loc=none, n_words=low, tot_silence=high, trim_dur=low</td>\n",
       "      <td>80.70</td>\n",
       "      <td>97.66</td>\n",
       "      <td>64.00</td>\n",
       "      <td>95.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>=</td>\n",
       "      <td>loc=none, trim_dur=high</td>\n",
       "      <td>96.17</td>\n",
       "      <td>99.62</td>\n",
       "      <td>98.82</td>\n",
       "      <td>98.82</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>&lt;</td>\n",
       "      <td>ageRange=41-65, n_words=high, speakRate=low</td>\n",
       "      <td>93.84</td>\n",
       "      <td>98.63</td>\n",
       "      <td>100.00</td>\n",
       "      <td>99.17</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Gain                                           Subgroup_str  \\\n",
       "0    >  loc=none, n_words=low, tot_silence=high, trim_dur=low   \n",
       "0    =                                loc=none, trim_dur=high   \n",
       "0    <            ageRange=41-65, n_words=high, speakRate=low   \n",
       "\n",
       "   wav2vec2-base_gender=female  hubert-base_gender=female  \\\n",
       "0                        80.70                      97.66   \n",
       "0                        96.17                      99.62   \n",
       "0                        93.84                      98.63   \n",
       "\n",
       "   wav2vec2-base_gender=male  hubert-base_gender=male  \n",
       "0                      64.00                    95.20  \n",
       "0                      98.82                    98.82  \n",
       "0                     100.00                    99.17  "
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Choose the top itemset for each gain value (i.e., >, <, =)\n",
    "comparison_top_1 = comparison_res_pr.loc[[0]]\n",
    "comparison_top_1 = comparison_top_1.sort_values(\"Gain\", ascending=False)\n",
    "comparison_top_1"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "327a8736",
   "metadata": {},
   "source": [
    "#### Shapley value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "eee769bd",
   "metadata": {},
   "outputs": [],
   "source": [
    "height = 0.45\n",
    "sizeFig = (6.5,2.5)\n",
    "labelsize = 12.3\n",
    "titlesize = 14\n",
    "pad = 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "c605f48b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 650x250 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Plot Shapley Values for the first just-retrieved data subgroup\n",
    "itemset = comparison_res.loc[0].iloc[0]['Subgroup']\n",
    "\n",
    "itemset_shaps = []\n",
    "for gender in genders:\n",
    "    itemset_0 = itemset.union([gender])\n",
    "    itemset_shaps.append(fp_divergence_difference.computeShapleyValue(itemset_0))\n",
    "                \n",
    "plotComparisonShapleyValues(abbreviateDict(itemset_shaps[0], abbreviations), \\\n",
    "                            abbreviateDict(itemset_shaps[1], abbreviations), \\\n",
    "                            sizeFig=sizeFig, sharedAxis=True, sort=\"paper\", \\\n",
    "                            height = height, saveFig=False, subcaption=False, \\\n",
    "                            labelsize=labelsize, pad=pad) "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e6463273",
   "metadata": {},
   "source": [
    "## Global Shapley value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "8913e491",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Compute the top-K global shapley values related to the gain in performance between Wav2Vec2-Base and Wav2Vec2-Large\n",
    "global_item_divergence_w2b_hb = fp_divergence_difference.computeGlobalShapleyValue()\n",
    "\n",
    "K = 15\n",
    "topK_global_w2b_hb = {k:v for k,v in global_item_divergence_w2b_hb.items() \\\n",
    "                        if k in sorted(global_item_divergence_w2b_hb, \n",
    "                        key=lambda x: abs(global_item_divergence_w2b_hb[x]))[::-1][:K]}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "027c4a3d",
   "metadata": {},
   "outputs": [],
   "source": [
    "from utils_analysis import plotShapleyValue\n",
    "\n",
    "sizeFig = (3.2,4)\n",
    "labelsize = 13\n",
    "titlesize = 13"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "d2f0a189",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 320x400 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Plot and Save the Image\n",
    "topK_global_w2b_hb_abbr = abbreviateDict(topK_global_w2b_hb, abbreviations )\n",
    "plotShapleyValue(shapley_values=topK_global_w2b_hb_abbr, \\\n",
    "                sizeFig=sizeFig, labelsize=labelsize, titlesize=titlesize, \\\n",
    "                title = r\"$\\tilde{\\Delta}^g_{gain} Wav2B - HubB$\", saveFig=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "009f55bf",
   "metadata": {},
   "outputs": [],
   "source": [
    "## Take the K Global Shapley Values for HubB-WavB starting from the top-K of WavL-WavB\n",
    "topK_global_w2b_hb_1_abbr = abbreviateDict({k:v for k,v in global_item_divergence_w2b_hb.items() \\\n",
    "                        if k in sorted(topK_global_w2b_w2l)}, abbreviations)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "4420bf1c",
   "metadata": {},
   "outputs": [],
   "source": [
    "from utils_analysis import plotMultipleSV\n",
    "\n",
    "name_fig = f\"plot_results/global_shapley_gain_w2b_w2l_hb_{K}.pdf\"\n",
    "labelsize = 19\n",
    "sizeFig = (5, 6.8)\n",
    "height = 0.35\n",
    "paper_exp = True\n",
    "\n",
    "legend = True\n",
    "if legend:\n",
    "    name_fig = f\"plot_results/global_shapley_gain_w2b_w2l_hb_{K}_legend.pdf\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "ec8e87a3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 500x680 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Plot and Save Image for Global Shapley Value Comparison\n",
    "\n",
    "plotMultipleSV(shapley_values_1=topK_global_w2b_w2l_abbr, \\\n",
    "            shapley_values_2=topK_global_w2b_hb_1_abbr, \\\n",
    "            paper_exp=paper_exp, labelsize=labelsize, \\\n",
    "            sizeFig=sizeFig, height=height,legend=legend, \\\n",
    "            nameFig=name_fig, saveFig=True)"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "name": "DivExplorer_FSC_IC.ipynb",
   "provenance": [],
   "toc_visible": true
  },
  "kernelspec": {
   "display_name": "Python 3.10.4 ('speech': conda)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.4"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "313.76837158203125px"
   },
   "toc_section_display": true,
   "toc_window_display": true
  },
  "vscode": {
   "interpreter": {
    "hash": "50f798c039f92e39594af06ec0119751541d975fa6ec3b2f5528645cd2e370ad"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}