06_embed_rand_hier.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "8816b366-5be4-4748-b769-4dc6f7d1735f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Modify the file 00_setup.py to define input/output file paths on your system\n",
    "# The information in 00_setup.py will be used across notebooks\n",
    "from importlib.machinery import SourceFileLoader\n",
    "setup = SourceFileLoader(\"setup\", \"./00_setup.py\").load_module()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a945de0c-f042-46b2-9d06-ab064af6c05b",
   "metadata": {},
   "source": [
    "## Neural Network With NAICS Embedding Layer + Random Unseen Codes\n",
    "This is like 04, but the training data is randomized to inject the unseen embedding code (1) into the data."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e8d6ba37-5510-46a9-8de1-e87d3dfb6712",
   "metadata": {},
   "source": [
    "Add NAICS embedding to model.  Only add the baseline NAICS (not higher levels in the hierarchy).  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e169c2e8-ba0f-488f-acf6-87d419bd4e99",
   "metadata": {},
   "source": [
    "Use the dataset with no missing values. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d47a7ec9-7d7b-47df-a902-b2f1e54f1135",
   "metadata": {},
   "source": [
    "*This script takes about 2 hours on my MacBook Air*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "fb42f9b4-99a8-4753-aacf-8722469be4c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "bf51170a-4487-4d10-b63d-c88f41b529f9",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "import importlib, pickle\n",
    "from sklearn import model_selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ae830c76-ee3d-4c70-870e-aec00a1fc79e",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "09dd973b-60a0-434b-9193-8c5fcec27dd5",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import os\n",
    "import re\n",
    "\n",
    "import keras\n",
    "from tensorflow.keras import layers, optimizers, losses, metrics, Model\n",
    "from IPython.display import display, HTML\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "9f1e4f2e-0e45-4e28-8e9c-212e0bb51199",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sba_nn.sba_nn import sbnn_metrics, sbnn_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "da3d6e2c-7884-41dc-8e8d-6ea5747c5e5a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<module 'sba_nn.sba_nn.sbnn_model' from '/Users/valeriecarey/Documents/projects/2024_05_blog_nn_rand/code/sba_nn/sba_nn/sbnn_model.py'>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "importlib.reload(sbnn_model)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3cbc00d2-bc83-42d9-83eb-65fcf306dccd",
   "metadata": {},
   "source": [
    "## Input Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "e014cd31-0f8d-42e3-82f9-b529eec8a3cc",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "sba_loans = pd.read_parquet(Path(setup.temp_path).joinpath('01_DATA_transformed_nomiss.parquet'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "8e65cb7c-792c-46ff-a530-0d5c64240d60",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(Path(setup.temp_path).joinpath('01_DATA_features.pkl'), 'rb') as fin:\n",
    "    imputer_features = pickle.load(fin)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "6f0ce28e-a80c-415b-b068-5c4d49370da8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'NAICS': 1170,\n",
       " 'NAICS_5': 764,\n",
       " 'NAICS_4': 345,\n",
       " 'NAICS_3': 107,\n",
       " 'NAICS_sector': 21}"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with open(Path(setup.temp_path).joinpath('01_DATA_naics_max_encodings.pkl'), 'rb') as fin:\n",
    "    naics_max_levels= pickle.load(fin)\n",
    "naics_max_levels"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4d4116da-d689-478a-9c5d-a330f231e615",
   "metadata": {},
   "source": [
    "##### Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "f98c6e41-38d0-4d5f-b821-e3837b8c2068",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['NoEmp', 'CreateJob', 'LowDoc', 'DisbursementGross', 'new_business', 'urban_flag', 'franchise_flag', 'missingindicator_LowDoc', 'missingindicator_new_business', 'missingindicator_urban_flag']\n"
     ]
    }
   ],
   "source": [
    "# Numeric features\n",
    "features_numeric = [f for f in imputer_features if 'NAICS' not in f]\n",
    "features_numeric_len = len(features_numeric)\n",
    "print(features_numeric)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "a74fa386-4011-45ee-bdb5-d671db9ffa91",
   "metadata": {},
   "outputs": [],
   "source": [
    "# NAICS feature - all levels except NAICS_5\n",
    "features_naics = ['NAICS', 'NAICS_4', 'NAICS_3', 'NAICS_sector']\n",
    "features_naics_max_levels  = [naics_max_levels[n] for n in features_naics]\n",
    "features_naics_emb_dim = [setup.nn_naics_embed_size_dict[n] for n in features_naics]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "867031ee-8b6e-424d-b497-8e19409ba1d9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1170, 345, 107, 21]\n",
      "[8, 4, 4, 2]\n"
     ]
    }
   ],
   "source": [
    "print(features_naics_max_levels)\n",
    "print(features_naics_emb_dim)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3da44371-2bb8-4b1f-ba5f-dd08571365d9",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Data Adjustment\n",
    "Inject \"1\" values into train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "9813c1ba-0668-40f0-82fa-f04d3b514e3a",
   "metadata": {},
   "outputs": [],
   "source": [
    "for i, n in enumerate(features_naics):\n",
    "    adj_index, _ = train_test_split(sba_loans[sba_loans['dset'] == 'train'].index, \n",
    "                                    train_size=0.10, random_state=645 + i)\n",
    "    sba_loans.loc[adj_index, n] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "61e7b587-cdf0-4074-810b-8462bb12d822",
   "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>NAICS</th>\n",
       "      <th>False</th>\n",
       "      <th>True</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dset</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>test</th>\n",
       "      <td>95552</td>\n",
       "      <td>57496</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>402164</td>\n",
       "      <td>44684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>val</th>\n",
       "      <td>95593</td>\n",
       "      <td>11</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "NAICS   False  True \n",
       "dset                \n",
       "test    95552  57496\n",
       "train  402164  44684\n",
       "val     95593     11"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Show unseen counts - expect some in train now!\n",
    "pd.crosstab(sba_loans['dset'], sba_loans['NAICS'] == 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b979396c-c822-4940-a292-6e07f1b54f58",
   "metadata": {},
   "source": [
    "##### Datasets for train, validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "43bca011-a59d-4496-9f03-87f4e350a253",
   "metadata": {},
   "outputs": [],
   "source": [
    "X = sba_loans[['dset', 'LoanNr_ChkDgt'] + features_numeric].set_index('LoanNr_ChkDgt').sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "aa162d01-ed73-4ae3-832e-4aeea09b70f4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training X: (446848, 10), y:(446848,)\n"
     ]
    }
   ],
   "source": [
    "X_train = X[X['dset'] == 'train'].drop(columns='dset')\n",
    "y_train = sba_loans[sba_loans['dset'] == 'train'].set_index('LoanNr_ChkDgt').sort_index()['target']\n",
    "print(f'training X: {X_train.shape}, y:{y_train.shape}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "b32b5fd3-6f6e-4aec-8e20-0e2b71a4814c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val X: (95604, 10), y:(95604,)\n"
     ]
    }
   ],
   "source": [
    "X_val = X[X['dset'] == 'val'].drop(columns='dset')\n",
    "y_val = sba_loans[sba_loans['dset'] == 'val'].set_index('LoanNr_ChkDgt').sort_index()['target']\n",
    "print(f'val X: {X_val.shape}, y:{y_val.shape}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "01b8ff96-159b-4283-b5fb-c4b5dbd277d5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.20230592953308507\n"
     ]
    }
   ],
   "source": [
    "base_thresh = y_train.mean()\n",
    "print(base_thresh)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "1b8ce4c9-6907-4b70-982a-b92a52b3159d",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_naics = [sba_loans[['dset', 'LoanNr_ChkDgt', n]].set_index('LoanNr_ChkDgt').sort_index() \\\n",
    "           for n in features_naics]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "a6f129e0-a4d0-4d93-9886-153c9803d403",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_naics_train = [Xn[Xn['dset'] == 'train'].drop(columns='dset') for Xn in X_naics]\n",
    "X_naics_val =  [Xn[Xn['dset'] == 'val'].drop(columns='dset') for Xn in X_naics]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a5c68623-4599-4472-a13a-9400d101633e",
   "metadata": {},
   "source": [
    "## Create, Fit Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "4d48ddfa-580c-4321-9917-0bd176b67d87",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-19 12:51:33.252550: I metal_plugin/src/device/metal_device.cc:1154] Metal device set to: Apple M1\n",
      "2024-05-19 12:51:33.252569: I metal_plugin/src/device/metal_device.cc:296] systemMemory: 16.00 GB\n",
      "2024-05-19 12:51:33.252574: I metal_plugin/src/device/metal_device.cc:313] maxCacheSize: 5.33 GB\n",
      "2024-05-19 12:51:33.252603: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:303] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support.\n",
      "2024-05-19 12:51:33.252615: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:269] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 0 MB memory) -> physical PluggableDevice (device: 0, name: METAL, pci bus id: <undefined>)\n"
     ]
    }
   ],
   "source": [
    "this_model = sbnn_model.create_emb_model(features_numeric_len,\n",
    "                                         features_naics_max_levels,\n",
    "                                         features_naics_emb_dim,\n",
    "                                         naics_embedding_names = features_naics,\n",
    "                                         hidden_size = setup.nn_layer_sizes,\n",
    "                                         activation='tanh',\n",
    "                                         lr=setup.nn_learning_rate,\n",
    "                                         opt_func = setup.nn_optimizer,\n",
    "                                         dropout = setup.nn_dropout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "47036621-ef13-4957-997b-f8ef86016f7f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                Output Shape                 Param #   Connected to                  \n",
      "==================================================================================================\n",
      " input_naics_00 (InputLayer  [(None, 1)]                  0         []                            \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " input_naics_01 (InputLayer  [(None, 1)]                  0         []                            \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " input_naics_02 (InputLayer  [(None, 1)]                  0         []                            \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " input_naics_03 (InputLayer  [(None, 1)]                  0         []                            \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " emb_naics_00 (Embedding)    (None, 1, 8)                 9360      ['input_naics_00[0][0]']      \n",
      "                                                                                                  \n",
      " emb_naics_01 (Embedding)    (None, 1, 4)                 1380      ['input_naics_01[0][0]']      \n",
      "                                                                                                  \n",
      " emb_naics_02 (Embedding)    (None, 1, 4)                 428       ['input_naics_02[0][0]']      \n",
      "                                                                                                  \n",
      " emb_naics_03 (Embedding)    (None, 1, 2)                 42        ['input_naics_03[0][0]']      \n",
      "                                                                                                  \n",
      " NAICS (Flatten)             (None, 8)                    0         ['emb_naics_00[0][0]']        \n",
      "                                                                                                  \n",
      " NAICS_4 (Flatten)           (None, 4)                    0         ['emb_naics_01[0][0]']        \n",
      "                                                                                                  \n",
      " NAICS_3 (Flatten)           (None, 4)                    0         ['emb_naics_02[0][0]']        \n",
      "                                                                                                  \n",
      " NAICS_sector (Flatten)      (None, 2)                    0         ['emb_naics_03[0][0]']        \n",
      "                                                                                                  \n",
      " input_numeric (InputLayer)  [(None, 10)]                 0         []                            \n",
      "                                                                                                  \n",
      " input_concat (Concatenate)  (None, 28)                   0         ['NAICS[0][0]',               \n",
      "                                                                     'NAICS_4[0][0]',             \n",
      "                                                                     'NAICS_3[0][0]',             \n",
      "                                                                     'NAICS_sector[0][0]',        \n",
      "                                                                     'input_numeric[0][0]']       \n",
      "                                                                                                  \n",
      " layer_00 (Dense)            (None, 128)                  3712      ['input_concat[0][0]']        \n",
      "                                                                                                  \n",
      " dropout_00 (Dropout)        (None, 128)                  0         ['layer_00[0][0]']            \n",
      "                                                                                                  \n",
      " layer_01 (Dense)            (None, 64)                   8256      ['dropout_00[0][0]']          \n",
      "                                                                                                  \n",
      " dropout_01 (Dropout)        (None, 64)                   0         ['layer_01[0][0]']            \n",
      "                                                                                                  \n",
      " output (Dense)              (None, 1)                    65        ['dropout_01[0][0]']          \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 23243 (90.79 KB)\n",
      "Trainable params: 23243 (90.79 KB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "this_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "54705746-b61f-4015-8940-bd594f8151a1",
   "metadata": {},
   "outputs": [],
   "source": [
    "callback = keras.callbacks.EarlyStopping(patience=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "31772016-3643-468b-8d65-c456168cad6e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-19 12:51:33.922842: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "13962/13964 [============================>.] - ETA: 0s - loss: 0.4632 - auc: 0.3341 - auc_1: 0.6983"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-19 12:54:10.862932: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "13964/13964 [==============================] - 176s 13ms/step - loss: 0.4632 - auc: 0.3341 - auc_1: 0.6983 - val_loss: 0.4550 - val_auc: 0.3704 - val_auc_1: 0.7244\n",
      "Epoch 2/20\n",
      "13964/13964 [==============================] - 189s 14ms/step - loss: 0.4548 - auc: 0.3578 - auc_1: 0.7160 - val_loss: 0.4510 - val_auc: 0.3789 - val_auc_1: 0.7285\n",
      "Epoch 3/20\n",
      "13964/13964 [==============================] - 216s 15ms/step - loss: 0.4525 - auc: 0.3654 - auc_1: 0.7208 - val_loss: 0.4502 - val_auc: 0.3834 - val_auc_1: 0.7305\n",
      "Epoch 4/20\n",
      "13964/13964 [==============================] - 229s 16ms/step - loss: 0.4514 - auc: 0.3690 - auc_1: 0.7232 - val_loss: 0.4498 - val_auc: 0.3840 - val_auc_1: 0.7304\n",
      "Epoch 5/20\n",
      "13964/13964 [==============================] - 227s 16ms/step - loss: 0.4505 - auc: 0.3721 - auc_1: 0.7249 - val_loss: 0.4498 - val_auc: 0.3843 - val_auc_1: 0.7309\n",
      "Epoch 6/20\n",
      "13964/13964 [==============================] - 231s 17ms/step - loss: 0.4498 - auc: 0.3748 - auc_1: 0.7263 - val_loss: 0.4500 - val_auc: 0.3846 - val_auc_1: 0.7302\n",
      "Epoch 7/20\n",
      "13964/13964 [==============================] - 246s 18ms/step - loss: 0.4492 - auc: 0.3769 - auc_1: 0.7273 - val_loss: 0.4500 - val_auc: 0.3844 - val_auc_1: 0.7307\n",
      "Epoch 8/20\n",
      "13964/13964 [==============================] - 204s 15ms/step - loss: 0.4490 - auc: 0.3775 - auc_1: 0.7279 - val_loss: 0.4502 - val_auc: 0.3845 - val_auc_1: 0.7303\n"
     ]
    }
   ],
   "source": [
    "this_history = this_model.fit([X_train] + X_naics_train, y_train,\n",
    "                              validation_data=([X_val] + X_naics_val, y_val),\n",
    "                              batch_size=setup.nn_batch_size,\n",
    "                              epochs=setup.nn_epochs,\n",
    "                              callbacks=[callback])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "bd448bf9-89d7-489a-9575-de9e2138e678",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<module 'sba_nn.sba_nn.sbnn_model' from '/Users/valeriecarey/Documents/projects/2024_05_blog_nn_rand/code/sba_nn/sba_nn/sbnn_model.py'>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "importlib.reload(sbnn_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "460c4b43-c673-4ee1-b070-d2d596ac4561",
   "metadata": {},
   "outputs": [],
   "source": [
    "this_history_df = sbnn_model.process_history(this_history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "ad4783f2-8d0b-4850-a45a-3a4be8e63318",
   "metadata": {},
   "outputs": [],
   "source": [
    "this_history_df.to_csv(Path(setup.temp_path).joinpath('06_REPORT_fit_history.csv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "73e89efd-75a9-4b27-81a8-13e12d5482be",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: >"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "this_history_df[['loss', 'val_loss']].plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "2342a31f-4891-4834-851a-b90b8b729c3c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: >"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "this_history_df[['auc', 'val_auc']].plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "cee80637-cabf-4974-b476-08c5fba165ff",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: >"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "this_history_df[['auc_roc', 'val_auc_roc']].plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "e184b0ca-1d31-4798-88a0-4e590b67062e",
   "metadata": {},
   "outputs": [],
   "source": [
    "this_model.save(Path(setup.temp_path).joinpath('06_DATA_model.keras'),save_format='tf')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d326c9ea-4759-4418-a64d-56c151049c25",
   "metadata": {},
   "source": [
    "## Predictions on all data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "d5c39a5a-dc19-4606-b744-34865bc6c5c5",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_naics = [sba_loans[['dset', 'LoanNr_ChkDgt', n]].set_index('LoanNr_ChkDgt').sort_index() \\\n",
    "           for n in features_naics]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "cd6910ee-5ffe-4def-9aa8-7dba76446b08",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   52/21735 [..............................] - ETA: 43s "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-19 13:20:12.135476: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "21735/21735 [==============================] - 43s 2ms/step\n"
     ]
    }
   ],
   "source": [
    "all_predictions = this_model.predict([X.drop(columns='dset')] +\n",
    "                                     [Xn.drop(columns='dset') for Xn in X_naics])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "98a2cce1-b7d3-4ebc-8869-80705297e145",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_predictions_df = pd.DataFrame(all_predictions, index=X.index) \\\n",
    "    .set_axis(['predict_prob'], axis=1) \\\n",
    "    .reset_index() \\\n",
    "    .merge(sba_loans[['target', 'LoanNr_ChkDgt', 'dset', 'dset_naics_holdout', 'NAICS']], \n",
    "           on='LoanNr_ChkDgt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "3fc7da9c-6bd2-4d1b-b77f-b62679825d5d",
   "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>predict_prob</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>predict_prob</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.325643</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>target</th>\n",
       "      <td>0.325643</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              predict_prob    target\n",
       "predict_prob      1.000000  0.325643\n",
       "target            0.325643  1.000000"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_predictions_df[['predict_prob', 'target']].corr(method='spearman')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e79afc4e-cb67-4064-9b19-dc52500baef2",
   "metadata": {},
   "source": [
    "##### Threshold Tune & Binary Predictions\n",
    "Using training probability predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "ccd0bc23-fbc0-40d0-8d8b-f2302405a875",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_pred_train = all_predictions_df[all_predictions_df['dset'] == 'train']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "df36d9e9-f35e-46b3-a5be-08081f998319",
   "metadata": {},
   "outputs": [],
   "source": [
    "thresh_tune_data = sbnn_metrics.get_f1_frame(all_pred_train['target'], \n",
    "                                        all_pred_train['predict_prob'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "de634156-caf6-4a52-b5d7-44b8e986aa1d",
   "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>thresh</th>\n",
       "      <th>f1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.22</td>\n",
       "      <td>0.454042</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.20</td>\n",
       "      <td>0.453799</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.24</td>\n",
       "      <td>0.450097</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    thresh        f1\n",
       "11    0.22  0.454042\n",
       "10    0.20  0.453799\n",
       "12    0.24  0.450097"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "thresh_tune_data.sort_values('f1', ascending=False, inplace=True)\n",
    "thresh_tune_data.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "c746b880-d781-4bc1-9df2-89211127c22a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.22"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_thresh = thresh_tune_data['thresh'].iloc[0]\n",
    "best_thresh"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c0ac5e2-be0d-42f0-960d-1f81b4a5d4e2",
   "metadata": {},
   "source": [
    "##### Append binary predictions to probability predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "641da72e-1bca-4fe7-96c4-6c029b521bd7",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_predictions_df['predict_bin'] = sbnn_metrics.get_binary_predictions(all_predictions_df['predict_prob'], best_thresh)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "9f6ecdf4-ac01-4c39-8315-e8583daa337e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "predict_bin\n",
       "0    0.599376\n",
       "1    0.400624\n",
       "Name: proportion, dtype: float64"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_predictions_df['predict_bin'].value_counts(normalize=True, dropna=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "cbfed830-7ff8-40f6-a691-6b894b26d6bc",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_predictions_df.to_parquet(Path(setup.temp_path).joinpath('06_DATA_predictions.parquet'))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eb0ecbd2-abff-46d8-8c70-f7bccd363b69",
   "metadata": {},
   "source": [
    "## Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "be4ceefe-c7a8-45bd-8573-2daefe99a1c2",
   "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>dset</th>\n",
       "      <th>accuracy_score</th>\n",
       "      <th>f1_score</th>\n",
       "      <th>precision_score</th>\n",
       "      <th>recall_score</th>\n",
       "      <th>average_precision_score</th>\n",
       "      <th>roc_auc_score</th>\n",
       "      <th>dset_naics_holdout</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>test</td>\n",
       "      <td>0.666850</td>\n",
       "      <td>0.449968</td>\n",
       "      <td>0.337323</td>\n",
       "      <td>0.675564</td>\n",
       "      <td>0.379916</td>\n",
       "      <td>0.730735</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>train</td>\n",
       "      <td>0.671248</td>\n",
       "      <td>0.454042</td>\n",
       "      <td>0.341883</td>\n",
       "      <td>0.675719</td>\n",
       "      <td>0.391884</td>\n",
       "      <td>0.735882</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>val</td>\n",
       "      <td>0.669690</td>\n",
       "      <td>0.451953</td>\n",
       "      <td>0.341454</td>\n",
       "      <td>0.668189</td>\n",
       "      <td>0.384567</td>\n",
       "      <td>0.730274</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>test</td>\n",
       "      <td>0.667152</td>\n",
       "      <td>0.446197</td>\n",
       "      <td>0.337033</td>\n",
       "      <td>0.659954</td>\n",
       "      <td>0.379570</td>\n",
       "      <td>0.726183</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>test</td>\n",
       "      <td>0.666348</td>\n",
       "      <td>0.456108</td>\n",
       "      <td>0.337785</td>\n",
       "      <td>0.702016</td>\n",
       "      <td>0.381281</td>\n",
       "      <td>0.738753</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    dset  accuracy_score  f1_score  precision_score  recall_score  \\\n",
       "0   test        0.666850  0.449968         0.337323      0.675564   \n",
       "1  train        0.671248  0.454042         0.341883      0.675719   \n",
       "2    val        0.669690  0.451953         0.341454      0.668189   \n",
       "0   test        0.667152  0.446197         0.337033      0.659954   \n",
       "1   test        0.666348  0.456108         0.337785      0.702016   \n",
       "\n",
       "   average_precision_score  roc_auc_score  dset_naics_holdout  \n",
       "0                 0.379916       0.730735                 NaN  \n",
       "1                 0.391884       0.735882                 NaN  \n",
       "2                 0.384567       0.730274                 NaN  \n",
       "0                 0.379570       0.726183                 0.0  \n",
       "1                 0.381281       0.738753                 1.0  "
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metrics_dset_df = all_predictions_df.groupby('dset') \\\n",
    "    .apply(lambda x: sbnn_metrics.dset_metrics(x.target, x.predict_bin, x.predict_prob)) \\\n",
    "    .reset_index()\n",
    "metrics_test_df = all_predictions_df[all_predictions_df['dset'] == 'test'] \\\n",
    "    .groupby(['dset', 'dset_naics_holdout']) \\\n",
    "    .apply(lambda x: sbnn_metrics.dset_metrics(x.target, x.predict_bin, x.predict_prob)) \\\n",
    "    .reset_index()\n",
    "metrics_df = pd.concat([metrics_dset_df, metrics_test_df])\n",
    "metrics_df.to_csv(Path(setup.temp_path).joinpath('06_REPORT_metrics.csv'), index=True)\n",
    "metrics_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "52e382b5-e9dc-47a1-9bf4-6d0b6b4a0c61",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tf_p39",
   "language": "python",
   "name": "tf_p39"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}