final_project_hierachical_clustering.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**What we do in this notebook:**\n",
    "\n",
    "Here we use Hierarchical clustering to cluster our song data into clusters and then recommend songs from the same cluster. The idea is that songs in the same cluster are more likely to be similar. Thus given, a playlist, we will recommend it songs that are in the same cluster most of the playlists' songs are but not in the playlist.\n",
    "\n",
    "Types of Clustering\n",
    "We used Agglomerative clustering (Bottom-up approach) instead of Divisive clustering since it is less complex and given how time consuming our calculations are, it will be less-expensive to compute while producing similar results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import json\n",
    "\n",
    "from sklearn.cluster import AgglomerativeClustering\n",
    "from scipy.cluster.hierarchy import dendrogram, linkage\n",
    "from scipy.spatial.distance import euclidean\n",
    "from collections import defaultdict\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.utils import resample\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "import seaborn as sns\n",
    "sns.set(style='whitegrid')\n",
    "pd.set_option('display.width', 1500)\n",
    "pd.set_option('display.max_columns', 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reading the dataframe to resuming working\n",
    "df = pd.read_csv(\"data/Sample_from_Million_Playlist.csv\")\n",
    "\n",
    "# Reading the data from saved file\n",
    "# This is the dataframe we will implement our mode\n",
    "df_spotify = pd.read_csv(\"data/100_Sample_MilPlay_Spotify.csv\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>duration_ms</th>\n",
       "      <th>danceability</th>\n",
       "      <th>energy</th>\n",
       "      <th>key</th>\n",
       "      <th>loudness</th>\n",
       "      <th>speechiness</th>\n",
       "      <th>acousticness</th>\n",
       "      <th>instrumentalness</th>\n",
       "      <th>liveness</th>\n",
       "      <th>valence</th>\n",
       "      <th>tempo</th>\n",
       "      <th>time_signature</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>199453</td>\n",
       "      <td>0.376</td>\n",
       "      <td>0.821</td>\n",
       "      <td>0</td>\n",
       "      <td>-3.974</td>\n",
       "      <td>0.1040</td>\n",
       "      <td>0.0785</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0823</td>\n",
       "      <td>0.331</td>\n",
       "      <td>120.462</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>260640</td>\n",
       "      <td>0.585</td>\n",
       "      <td>0.421</td>\n",
       "      <td>7</td>\n",
       "      <td>-9.593</td>\n",
       "      <td>0.0707</td>\n",
       "      <td>0.1060</td>\n",
       "      <td>0.000010</td>\n",
       "      <td>0.1400</td>\n",
       "      <td>0.240</td>\n",
       "      <td>132.986</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>239320</td>\n",
       "      <td>0.504</td>\n",
       "      <td>0.705</td>\n",
       "      <td>6</td>\n",
       "      <td>-8.205</td>\n",
       "      <td>0.3640</td>\n",
       "      <td>0.0752</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.1280</td>\n",
       "      <td>0.584</td>\n",
       "      <td>175.483</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>206240</td>\n",
       "      <td>0.771</td>\n",
       "      <td>0.647</td>\n",
       "      <td>5</td>\n",
       "      <td>-5.127</td>\n",
       "      <td>0.3760</td>\n",
       "      <td>0.2940</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.3700</td>\n",
       "      <td>0.678</td>\n",
       "      <td>155.914</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>235653</td>\n",
       "      <td>0.480</td>\n",
       "      <td>0.682</td>\n",
       "      <td>7</td>\n",
       "      <td>-4.940</td>\n",
       "      <td>0.1300</td>\n",
       "      <td>0.0696</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0463</td>\n",
       "      <td>0.506</td>\n",
       "      <td>167.939</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>226013</td>\n",
       "      <td>0.511</td>\n",
       "      <td>0.596</td>\n",
       "      <td>0</td>\n",
       "      <td>-6.560</td>\n",
       "      <td>0.2240</td>\n",
       "      <td>0.5300</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0798</td>\n",
       "      <td>0.554</td>\n",
       "      <td>158.063</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>261160</td>\n",
       "      <td>0.422</td>\n",
       "      <td>0.712</td>\n",
       "      <td>11</td>\n",
       "      <td>-5.907</td>\n",
       "      <td>0.1000</td>\n",
       "      <td>0.2730</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0510</td>\n",
       "      <td>0.471</td>\n",
       "      <td>78.454</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>240147</td>\n",
       "      <td>0.418</td>\n",
       "      <td>0.111</td>\n",
       "      <td>1</td>\n",
       "      <td>-14.848</td>\n",
       "      <td>0.0389</td>\n",
       "      <td>0.9530</td>\n",
       "      <td>0.003420</td>\n",
       "      <td>0.0982</td>\n",
       "      <td>0.228</td>\n",
       "      <td>82.508</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>219214</td>\n",
       "      <td>0.377</td>\n",
       "      <td>0.638</td>\n",
       "      <td>4</td>\n",
       "      <td>-5.754</td>\n",
       "      <td>0.0545</td>\n",
       "      <td>0.2130</td>\n",
       "      <td>0.000080</td>\n",
       "      <td>0.1200</td>\n",
       "      <td>0.369</td>\n",
       "      <td>92.644</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>283466</td>\n",
       "      <td>0.395</td>\n",
       "      <td>0.676</td>\n",
       "      <td>0</td>\n",
       "      <td>-5.460</td>\n",
       "      <td>0.0498</td>\n",
       "      <td>0.6080</td>\n",
       "      <td>0.000049</td>\n",
       "      <td>0.1170</td>\n",
       "      <td>0.315</td>\n",
       "      <td>142.929</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   duration_ms  danceability  energy  key  loudness  speechiness  acousticness  instrumentalness  liveness  valence    tempo  time_signature\n",
       "0       199453         0.376   0.821    0    -3.974       0.1040        0.0785          0.000000    0.0823    0.331  120.462               4\n",
       "1       260640         0.585   0.421    7    -9.593       0.0707        0.1060          0.000010    0.1400    0.240  132.986               3\n",
       "2       239320         0.504   0.705    6    -8.205       0.3640        0.0752          0.000000    0.1280    0.584  175.483               4\n",
       "3       206240         0.771   0.647    5    -5.127       0.3760        0.2940          0.000000    0.3700    0.678  155.914               4\n",
       "4       235653         0.480   0.682    7    -4.940       0.1300        0.0696          0.000000    0.0463    0.506  167.939               3\n",
       "5       226013         0.511   0.596    0    -6.560       0.2240        0.5300          0.000000    0.0798    0.554  158.063               4\n",
       "6       261160         0.422   0.712   11    -5.907       0.1000        0.2730          0.000000    0.0510    0.471   78.454               4\n",
       "7       240147         0.418   0.111    1   -14.848       0.0389        0.9530          0.003420    0.0982    0.228   82.508               4\n",
       "8       219214         0.377   0.638    4    -5.754       0.0545        0.2130          0.000080    0.1200    0.369   92.644               4\n",
       "9       283466         0.395   0.676    0    -5.460       0.0498        0.6080          0.000049    0.1170    0.315  142.929               5"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Cleaning the data frame and removing features we don't need\n",
    "df_spotify = df_spotify.drop_duplicates(['track_uri'])\n",
    "df_spotify_tracks = df_spotify['track_uri']\n",
    "df_spotify_track_names = df_spotify['track_name']\n",
    "\n",
    "# Cleaning the data frame and removing features we don't need\n",
    "df_spotify = df_spotify.drop_duplicates(['track_uri']).reset_index()\n",
    "df_spotify_tracks = df_spotify['track_uri']\n",
    "df_spotify_track_names = df_spotify['track_name']\n",
    "\n",
    "# We will be using the clean df to implement our model\n",
    "df_spotify_clean = df_spotify.drop(columns=['index', 'pid','pos','artist_name','track_uri','artist_uri','track_name','album_uri','album_name','type','id','uri','track_href','analysis_url','duration_ms.1','mode'])\n",
    "df_spotify_clean.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Standardizing the data\n",
    "scaler = MinMaxScaler()\n",
    "scaler.fit(df_spotify_clean)\n",
    "df_spotify_clean_scaled = scaler.transform(df_spotify_clean)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Improved model\n",
    "\n",
    "def generate_playlist_hierarchical_cluster(playlist, df_scaled_features, info_df, cluster, random_state = 10):\n",
    "    \n",
    "    '''\n",
    "    Input:\n",
    "    playlist = Playlist based on which to provide recommendations \n",
    "    df_scaled_features = df with all the scaled audio features of our songs\n",
    "    info_df = The df with all the songs and track info and audio features\n",
    "    cluster = agglomerative clustering model.\n",
    "    random_state = Random state for k-clustering \n",
    "    \n",
    "    Output:\n",
    "    recs = A df that will be the recommended playlist\n",
    "    '''\n",
    "    pid = list(info_df[info_df['track_uri'].isin(playlist)]['pid'].value_counts().index)[0]\n",
    "    df_k = pd.DataFrame(df_scaled_features.copy())\n",
    "    \n",
    "    # get label\n",
    "    label = cluster.labels_\n",
    "    # Assigning labels to each song\n",
    "    df_k['cluster_label'] = label\n",
    "    \n",
    "    # For a given playlist, checking which cluster label the playlist had the highest\n",
    "    mode_cluster = list(df_k[info_df['pid'] == pid]['cluster_label'].value_counts().index)[0]\n",
    "    \n",
    "    # Finding songs not on our playlist by with same cluster label\n",
    "    our_df = info_df[~(info_df['track_uri'].isin(playlist)) & (df_k['cluster_label'] == mode_cluster)]\n",
    "\n",
    "    # Then among those songs, taking a single song from each artist \n",
    "    # The loop starts with the artist with the highest songs and then goes down and repeats\n",
    "    # We do this until our recommendation is complete\n",
    "    \n",
    "    artist_names = list(info_df[df_k['cluster_label'] == mode_cluster]['artist_name'].value_counts().index)\n",
    "    recs = []\n",
    "    counter = 0\n",
    "    while (len(recs) < 2*len(playlist)) and (counter < len(our_df[our_df['artist_name'] == artist_names[0]]['track_uri'])):    \n",
    "        for artist in artist_names:\n",
    "                songs = our_df[our_df['artist_name'] == artist]['track_uri']\n",
    "                if len(songs) > counter + 1:\n",
    "                    recs.append(list(songs)[counter])\n",
    "        counter += 1\n",
    "    return info_df[info_df['track_uri'].isin(recs)]['track_name'].iloc[0:len(playlist)]  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# This is one of the metrics that can be used to understand the precision of the recommendation\n",
    "def r_precision(prediction, validation):\n",
    "    score = np.sum(validation.isin(prediction))/validation.shape[0]\n",
    "    return score\n",
    "\n",
    "### NDCG Code Source: https://gist.github.com/bwhite/3726239\n",
    "def dcg_at_k(r, k, method=0):\n",
    "    r = np.asfarray(r)[:k]\n",
    "    if r.size:\n",
    "        if method == 0:\n",
    "            return r[0] + np.sum(r[1:] / np.log2(np.arange(2, r.size + 1)))\n",
    "        elif method == 1:\n",
    "            return np.sum(r / np.log2(np.arange(2, r.size + 2)))\n",
    "        else:\n",
    "            raise ValueError('method must be 0 or 1.')\n",
    "    return 0.\n",
    "\n",
    "\n",
    "def ndcg_at_k(r, k, method=0):\n",
    "    dcg_max = dcg_at_k(sorted(r, reverse=True), k, method)\n",
    "    if not dcg_max:\n",
    "        return 0.\n",
    "    return dcg_at_k(r, k, method) / dcg_max"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of clusters: 1\n",
      "Number of clusters: 2\n",
      "Number of clusters: 3\n",
      "Number of clusters: 4\n",
      "Number of clusters: 5\n",
      "Number of clusters: 6\n",
      "Number of clusters: 7\n",
      "Number of clusters: 8\n",
      "Number of clusters: 9\n"
     ]
    }
   ],
   "source": [
    "# Testing model\n",
    "\n",
    "# Choosing pids with at least 50 songs such that our model gets at least 25 songs to work with\n",
    "# The average playlist in our data set contains 52 songs\n",
    "# So, if we take 100 songs at least and do a .5 train-test split\n",
    "# The model will produce 50 recommendations and get tested on a 25 song playlist on average which is reasonable\n",
    "\n",
    "pids = list(df_spotify['pid'].value_counts()[df_spotify['pid'].value_counts() > 50].index)[0:30]\n",
    "\n",
    "mean_rprec = []\n",
    "mean_ndcgs = []\n",
    "\n",
    "for num_clusters in range(1,10):\n",
    "    r_precisions = []\n",
    "    ndcgs = []\n",
    "    cluster = AgglomerativeClustering(n_clusters=num_clusters, affinity='euclidean', linkage='ward')\n",
    "    df_k = pd.DataFrame(df_spotify_clean_scaled.copy())\n",
    "\n",
    "    # fit clusters\n",
    "    cluster.fit_predict(df_k)\n",
    "    for pid in pids:\n",
    "        train, test = train_test_split(df_spotify[df_spotify['pid'] == pid], test_size=0.50, random_state = 24)\n",
    "        prediction = generate_playlist_hierarchical_cluster(train['track_uri'], df_spotify_clean_scaled, df_spotify, cluster)\n",
    "        r = r_precision(prediction, test['track_name'])\n",
    "        r_precisions.append(r)\n",
    "\n",
    "        rx = np.zeros(len((prediction)))\n",
    "        for i, p in enumerate(prediction):\n",
    "            if np.any(test['track_name'].isin([p])):              \n",
    "                rx[i] = 1\n",
    "        ndcgs.append(ndcg_at_k(rx, len(rx)))\n",
    "    \n",
    "    print(\"Number of clusters: \" + str(num_clusters))\n",
    "    mean_rprec.append(np.mean(r_precisions))\n",
    "    mean_ndcgs.append(np.mean(ndcgs))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(np.arange(1, 10), mean_rprec)\n",
    "plt.title(\"R-precision scores vs. clusters for Hierarchical Clustering Model\")\n",
    "plt.xlabel(\"Number of clusters\")\n",
    "plt.ylabel(\"R-precision score\")\n",
    "plt.savefig(\"data/HC_R.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(np.arange(1, 10), mean_ndcgs)\n",
    "plt.title(\"NDCG scores vs. clusters for Hierarchical Clustering Model\")\n",
    "plt.xlabel(\"Number of clusters\")\n",
    "plt.ylabel(\"NDCG score\")\n",
    "plt.savefig(\"data/HC_NDCG.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best cluster:4\n",
      "r precision for that cluster: 0.01961842753064936\n",
      "NDCGs for that cluster: 0.0834216104106089\n"
     ]
    }
   ],
   "source": [
    "best_cluster = np.argmax(np.array(mean_rprec)+np.array(mean_ndcgs))\n",
    "print(\"Best cluster:{}\". format(best_cluster + 1))\n",
    "print(\"r precision for that cluster: {}\".format(mean_rprec[best_cluster]))\n",
    "print(\"NDCGs for that cluster: {}\".format(mean_ndcgs[best_cluster]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}