FIN5622_Project_5_Neil_Yejjey_tuk12930.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<h2>FIN5622 Project #5\n",
    "    <p class=\"text-primary\">Loan Default Prediction with Neural Networks</p>\n",
    "</h2>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-success\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:black\"><b>Project Introduction</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    In this project we will build a neural network model to predict borrower defaults on a set of loan performance data.\n",
    "    </div>\n",
    "</div>\n",
    "\n",
    "<div class = \"panel panel-info\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:black\"><b>Instructions</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    Please code you solutions directly in this Notebook under each task in the cells provided below. Tasks are intended to be done in order with results from one task informing work in later tasks. Include all code used to produce your results, and support your answers with textual explanations if neccessary. <b>Any written text must be typed in a markdown cell to receive credit!</b> You may feel free to add additional Code and/or Markdown cells as needed. Code must be written using <b>Python 3</b>.\n",
    "    <br><br>Please order the cells in your Notebook in the order in which they are intended to run! It is expected that your Notebook is able to run in sequence from start to finish without any errors (before submitting check by: <code>Kernel > Restart & Run All</code>).\n",
    "    <br><br>Once ready, please submit your assignment as an <b>.ipynb file</b> (<code>File > Download as > Notebook</code>). You must name your files in the following format: <b><code>FIN5622_Project_5_Richard_Lee_tuc12345.ipynb</code></b>.\n",
    "    </div>\n",
    "</div>\n",
    "\n",
    "---\n",
    "<div class = \"alert alert-block alert-warning\">\n",
    "    <b>Note</b>: Import all libraries that you plan on using in the code block below.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TensorFlow version: 2.8.0\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import roc_curve, roc_auc_score\n",
    "\n",
    "import tensorflow as tf\n",
    "print(\"TensorFlow version:\", tf.__version__) #Check Tensorflow version is at least 2.8!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-primary\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:white\"><b>Task 1</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    Read in the dataset \"loan_performance_data_2016.csv\" into a dataframe called <code>loans</code> using the pandas <code>pd.read_csv()</code> function. Check that there are 405,180 loans present in the dataframe.\n",
    "    <br><br>\n",
    "    Note that there are 16 feature columns and 1 target column (i.e. \"loan_default\") in this dataset. Please review the file \"Loan_Data_Dictionary.xlsx\" for definitions on each column.\n",
    "    <p style=\"text-align:right;\">[0pts]</p>\n",
    "    </div>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>loan_amnt</th>\n",
       "      <th>loan_term</th>\n",
       "      <th>int_rate</th>\n",
       "      <th>fico</th>\n",
       "      <th>dti</th>\n",
       "      <th>annual_income</th>\n",
       "      <th>income_verification</th>\n",
       "      <th>home_mortgage</th>\n",
       "      <th>home_own</th>\n",
       "      <th>home_rent</th>\n",
       "      <th>debt_consolidation</th>\n",
       "      <th>coborrower</th>\n",
       "      <th>num_accounts</th>\n",
       "      <th>revol_bal</th>\n",
       "      <th>revol_util</th>\n",
       "      <th>derogatory_records</th>\n",
       "      <th>loan_default</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7.200</td>\n",
       "      <td>3.0</td>\n",
       "      <td>5.32</td>\n",
       "      <td>750</td>\n",
       "      <td>19.05</td>\n",
       "      <td>49.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>11</td>\n",
       "      <td>9.309</td>\n",
       "      <td>18.4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>12.000</td>\n",
       "      <td>5.0</td>\n",
       "      <td>11.99</td>\n",
       "      <td>670</td>\n",
       "      <td>6.59</td>\n",
       "      <td>49.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>10</td>\n",
       "      <td>12.152</td>\n",
       "      <td>50.6</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>11.425</td>\n",
       "      <td>3.0</td>\n",
       "      <td>19.53</td>\n",
       "      <td>730</td>\n",
       "      <td>35.56</td>\n",
       "      <td>26.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>11</td>\n",
       "      <td>2.096</td>\n",
       "      <td>7.2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>17.500</td>\n",
       "      <td>3.0</td>\n",
       "      <td>13.67</td>\n",
       "      <td>665</td>\n",
       "      <td>38.29</td>\n",
       "      <td>46.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>33</td>\n",
       "      <td>18.847</td>\n",
       "      <td>62.6</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>22.875</td>\n",
       "      <td>5.0</td>\n",
       "      <td>17.27</td>\n",
       "      <td>670</td>\n",
       "      <td>11.75</td>\n",
       "      <td>65.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>6</td>\n",
       "      <td>25.200</td>\n",
       "      <td>90.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>405175</th>\n",
       "      <td>16.000</td>\n",
       "      <td>5.0</td>\n",
       "      <td>12.79</td>\n",
       "      <td>665</td>\n",
       "      <td>12.25</td>\n",
       "      <td>150.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>12</td>\n",
       "      <td>7.700</td>\n",
       "      <td>55.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>405176</th>\n",
       "      <td>24.000</td>\n",
       "      <td>5.0</td>\n",
       "      <td>12.79</td>\n",
       "      <td>665</td>\n",
       "      <td>19.61</td>\n",
       "      <td>95.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>49.431</td>\n",
       "      <td>84.4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>405177</th>\n",
       "      <td>40.000</td>\n",
       "      <td>5.0</td>\n",
       "      <td>10.49</td>\n",
       "      <td>705</td>\n",
       "      <td>12.75</td>\n",
       "      <td>227.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>8.633</td>\n",
       "      <td>64.9</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>405178</th>\n",
       "      <td>24.000</td>\n",
       "      <td>5.0</td>\n",
       "      <td>14.49</td>\n",
       "      <td>660</td>\n",
       "      <td>18.30</td>\n",
       "      <td>110.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>10</td>\n",
       "      <td>17.641</td>\n",
       "      <td>68.1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>405179</th>\n",
       "      <td>14.000</td>\n",
       "      <td>5.0</td>\n",
       "      <td>14.49</td>\n",
       "      <td>660</td>\n",
       "      <td>23.36</td>\n",
       "      <td>95.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>7.662</td>\n",
       "      <td>54.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>405180 rows × 17 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        loan_amnt  loan_term  int_rate  fico    dti  annual_income  \\\n",
       "0           7.200        3.0      5.32   750  19.05           49.0   \n",
       "1          12.000        5.0     11.99   670   6.59           49.0   \n",
       "2          11.425        3.0     19.53   730  35.56           26.0   \n",
       "3          17.500        3.0     13.67   665  38.29           46.0   \n",
       "4          22.875        5.0     17.27   670  11.75           65.0   \n",
       "...           ...        ...       ...   ...    ...            ...   \n",
       "405175     16.000        5.0     12.79   665  12.25          150.0   \n",
       "405176     24.000        5.0     12.79   665  19.61           95.0   \n",
       "405177     40.000        5.0     10.49   705  12.75          227.0   \n",
       "405178     24.000        5.0     14.49   660  18.30          110.0   \n",
       "405179     14.000        5.0     14.49   660  23.36           95.0   \n",
       "\n",
       "        income_verification  home_mortgage  home_own  home_rent  \\\n",
       "0                         0              1         0          0   \n",
       "1                         0              1         0          0   \n",
       "2                         0              0         0          1   \n",
       "3                         0              0         0          1   \n",
       "4                         0              0         0          1   \n",
       "...                     ...            ...       ...        ...   \n",
       "405175                    0              0         0          1   \n",
       "405176                    0              1         0          0   \n",
       "405177                    1              1         0          0   \n",
       "405178                    0              0         0          1   \n",
       "405179                    1              1         0          0   \n",
       "\n",
       "        debt_consolidation  coborrower  num_accounts  revol_bal  revol_util  \\\n",
       "0                        1           0            11      9.309        18.4   \n",
       "1                        0           0            10     12.152        50.6   \n",
       "2                        0           0            11      2.096         7.2   \n",
       "3                        1           0            33     18.847        62.6   \n",
       "4                        1           0             6     25.200        90.0   \n",
       "...                    ...         ...           ...        ...         ...   \n",
       "405175                   0           0            12      7.700        55.0   \n",
       "405176                   0           0             5     49.431        84.4   \n",
       "405177                   0           0             5      8.633        64.9   \n",
       "405178                   1           0            10     17.641        68.1   \n",
       "405179                   0           0             8      7.662        54.0   \n",
       "\n",
       "        derogatory_records  loan_default  \n",
       "0                        0             0  \n",
       "1                        0             1  \n",
       "2                        0             0  \n",
       "3                        0             0  \n",
       "4                        0             0  \n",
       "...                    ...           ...  \n",
       "405175                   1             0  \n",
       "405176                   0             0  \n",
       "405177                   0             0  \n",
       "405178                   1             1  \n",
       "405179                   0             0  \n",
       "\n",
       "[405180 rows x 17 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loans = pd.read_csv('loan_performance_data_2016.csv')\n",
    "loan_target = ['loan_default']\n",
    "loan_features = loans.drop('loan_default', axis = 1).columns.tolist()\n",
    "loans"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-primary\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:white\"><b>Task 2</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    Use the function <code>train_test_split()</code> from the <code>sklearn.model_selection</code> module to split <code>loans</code> into training and test datasets. Provide the function with 2 separate numpy arrays, storing the features and target separately (like we did in class). Additionally, set the following optional paramters for the function to the values listed below:\n",
    "    <ul>\n",
    "        <li><code>test_size = .2</code></li>\n",
    "        <li><code>random_state = 5622</code></li>\n",
    "    </ul>\n",
    "    <br>\n",
    "    The function should return 4 outputs, representing the features and target for the training and test datasets, and store them as numpy arrays in variables called <code>X_train</code>, <code>X_test</code>, <code>y_train</code>, and <code>y_test</code>. The training data should contain 80%, while the test data should contain 20%, of the original <code>loans</code> dataset. \n",
    "    <br><br>\n",
    "    <b>Convert the four numpy arrays into tensors, using the function <code>tf.convert_to_tensor()</code> from the tensorflow library.</b>\n",
    "    <br><br>\n",
    "    Check that each tensor shoud have the following shapes: \n",
    "    <ul>\n",
    "        <li><code>X_train.shape = (324144, 16)</code></li>\n",
    "        <li><code>X_test.shape = (81036, 16)</code></li>\n",
    "        <li><code>y_train.shape = (324144,)</code></li>\n",
    "        <li><code>y_test.shape = (81036,)</code></li>\n",
    "    </ul>\n",
    "    <p style=\"text-align:right;\">[15pts]</p>\n",
    "    </div>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    loans[loan_features], loans[loan_target], test_size = 0.20, random_state = 5622)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = tf.convert_to_tensor(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test = tf.convert_to_tensor(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = tf.convert_to_tensor(y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test = tf.convert_to_tensor(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TensorShape([324144, 16])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TensorShape([81036, 16])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TensorShape([324144, 1])"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TensorShape([81036, 1])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test.shape "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-primary\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:white\"><b>Task 3</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    Create a neural network model with the following architecture (recall this is done using the tensor flow function <code>tf.keras.models.Sequential()</code>), and store this in a variable called <code>nnet_model_1</code>:\n",
    "    <ul>\n",
    "        <li>A normalization layer with means and standard deviations populated by the training data <code>X_train</code>. The tensorflow function to create the normalization layer is <code>tf.keras.layers.Normalization(axis = -1)</code> and the <code>.adapt()</code> method.</li>\n",
    "        <li>A hidden layer with 64 nodes and ReLU activation function. The tensorflow function to create a fully connected layer is <code>tf.keras.layers.Dense()</code> with <code>activation = 'relu'</code>.</li>\n",
    "        <li>A second hidden layer with 128 nodes and ReLU activation function.</li>\n",
    "        <li>An output layer with 1 node, since we are only interested in predicting the probability of loan defaults. We will not apply the sigmoid activation function yet, until after training is complete.</li>\n",
    "    </ul>\n",
    "    <p style=\"text-align:right;\">[15pts]</p>\n",
    "    </div>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "normalization_layer = tf.keras.layers.Normalization(axis = -1)\n",
    "normalization_layer.adapt(X_train)\n",
    "\n",
    "nnet_model_1 = tf.keras.models.Sequential([\n",
    "    normalization_layer,\n",
    "    tf.keras.layers.Dense(64, activation = 'relu'),\n",
    "    tf.keras.layers.Dropout(0.05),\n",
    "    tf.keras.layers.Dense(128, activation = 'relu'),\n",
    "    tf.keras.layers.Dropout(0.05),\n",
    "    tf.keras.layers.Dense(1),\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-primary\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:white\"><b>Task 4</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    Configure the tensorflow model compilier for <code>nnet_model_1</code>, which will facilitate the training of our nueral network. Recall that this is done by calling the <code>.compile()</code> method of the neural network model object. \n",
    "    <br><br>\n",
    "    The two parameters we will need to configure in are:\n",
    "    <ul>\n",
    "        <li><code>loss</code>: Since we are training a classification model to make binary class predictions, the tensorflow loss function to use here is <code>tf.keras.losses.BinaryCrossentropy(from_logits = True)</code> (Recall that this is equivalent to log-likelihood). </li>\n",
    "        <li><code>optimizer</code>: Set Stochastic Gradient Decent (SGD) as the optimizer (tensorflow function is <code>ts.keras.optimizers.SGD()</code>. Set the learning rate equal to 0.1.</li>\n",
    "    </ul>\n",
    "    <p style=\"text-align:right;\">[20pts]</p>\n",
    "    </div>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_fx = tf.keras.losses.BinaryCrossentropy(from_logits = True)\n",
    "opt = tf.keras.optimizers.SGD(learning_rate = .1)\n",
    "\n",
    "nnet_model_1.compile(optimizer = opt, loss = loss_fx)"
   ]
  },
  {
   "attachments": {
    "Task_5_Plot-2.PNG": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-primary\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:white\"><b>Task 5</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "     Train the neural network. Recall that this is done by calling the <code>.fit()</code> method of the neural network model object. Provide the method with the training data (<code>X_train</code> and <code>y_train</code>), as well as the following additional parameters:\n",
    "    <ul>\n",
    "        <li><code>validation_data = (X_test, y_test)</code>: this allows for the evaluation of model performance on the test data during training.</li>\n",
    "        <li><code>epochs = 10</code>: trains the model on 10 passes of the training data.</li>\n",
    "        <li><code>shuffle = True</code>: shuffles the training data in between each epoch.</li>\n",
    "        <li><code>batch_size = 100</code>: sets the batch size to 100 samples under each iteration.</li>\n",
    "    </ul>\n",
    "    Plot both the training loss and test loss on a single line plot. Label the axes and provide a legend. Your plot should look similar to below:\n",
    "\n",
    "![Task_5_Plot-2.PNG](attachment:Task_5_Plot-2.PNG)\n",
    "    <br>\n",
    "    Based on the plot of training and test loss, did the neural network successful learn from the training data?\n",
    "    <p style=\"text-align:right;\">[20pts]</p>\n",
    "    </div>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "3242/3242 [==============================] - 14s 3ms/step - loss: 0.4516 - val_loss: 0.4453\n",
      "Epoch 2/10\n",
      "3242/3242 [==============================] - 11s 3ms/step - loss: 0.4487 - val_loss: 0.4447\n",
      "Epoch 3/10\n",
      "3242/3242 [==============================] - 11s 3ms/step - loss: 0.4481 - val_loss: 0.4445\n",
      "Epoch 4/10\n",
      "3242/3242 [==============================] - 9s 3ms/step - loss: 0.4477 - val_loss: 0.4447\n",
      "Epoch 5/10\n",
      "3242/3242 [==============================] - 8s 3ms/step - loss: 0.4474 - val_loss: 0.4442\n",
      "Epoch 6/10\n",
      "3242/3242 [==============================] - 8s 3ms/step - loss: 0.4472 - val_loss: 0.4447\n",
      "Epoch 7/10\n",
      "3242/3242 [==============================] - 8s 3ms/step - loss: 0.4469 - val_loss: 0.4441\n",
      "Epoch 8/10\n",
      "3242/3242 [==============================] - 9s 3ms/step - loss: 0.4467 - val_loss: 0.4444\n",
      "Epoch 9/10\n",
      "3242/3242 [==============================] - 8s 3ms/step - loss: 0.4468 - val_loss: 0.4443\n",
      "Epoch 10/10\n",
      "3242/3242 [==============================] - 8s 3ms/step - loss: 0.4465 - val_loss: 0.4442\n"
     ]
    }
   ],
   "source": [
    "history = nnet_model_1.fit(X_train, y_train, validation_data = (X_test, y_test), epochs = 10,\n",
    "                          shuffle = True, batch_size = 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'], label = 'loss')\n",
    "plt.plot(history.history['val_loss'], label = 'val_loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Based on the plot of training and test loss, the neural network was successful in learning from the training data as the loss and val_loss values conitnue to get closer to one another with each successive epoch. \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-primary\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:white\"><b>Task 6</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    Calculate and store the predictions from the neural network <code>nnet_model_1</code> on the test dataset <code>X_test</code> in a variable called <code>pred_1</code>. \n",
    "    <br><br>\n",
    "    Because the last layer of the neural network <code>nnet_model_1</code> is still a dense layer, the outputs of the model are still logits (linear combination of the output layer weights and last hidden layer features). Transform <code>pred_1</code> into probabilities using the tensorflow function <code>tf.nn.sigmoid()</code>. <b>Store this as a numpy array in a variable called <code>prob_1</code>.</b>\n",
    "    <br><br>\n",
    "    Recall, use the <code>.numpy()</code> method to turn tensors into numpy arrays.\n",
    "    <p style=\"text-align:right;\">[15pts]</p>\n",
    "    </div>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.2271218 ],\n",
       "       [0.22786453],\n",
       "       [0.31272572],\n",
       "       ...,\n",
       "       [0.10830709],\n",
       "       [0.23444769],\n",
       "       [0.18252414]], dtype=float32)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_1 = nnet_model_1(X_test)\n",
    "prob_1 = tf.nn.sigmoid(pred_1).numpy()\n",
    "prob_1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Your written response here*\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-primary\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:white\"><b>Task 7</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    Calculate the ROC curve and the AUC for <code>probs_1</code>. Recall that the ROC curve can be calculated using the function <code>roc_curve()</code>, and the AUC can be calculated using the function <code>roc_auc_score()</code>. Both functions are from the <code>sklearn.metrics</code> module.\n",
    "    <br><br>\n",
    "    Plot the ROC curve on a line plot. Recall that the ROC curve should be a upward-sloping and concaved curve above the 45-degree line. Provide axis labels and a title for the plot.\n",
    "    <p style=\"text-align:right;\">[15pts]</p>\n",
    "    </div>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fpr, tpr, th = roc_curve(y_test, prob_1)\n",
    "\n",
    "plt.plot(fpr, tpr)\n",
    "plt.plot(fpr, fpr, linestyle = '--', color = 'black')\n",
    "plt.ylabel('TPR')\n",
    "plt.xlabel('FPR')\n",
    "plt.title('ROC')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7190588091599402"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y_test, prob_1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.20025562, 0.1950614541685177)"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prob_1.mean(), y_test.numpy().mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": false,
    "editable": false
   },
   "source": [
    "<div class = \"panel panel-primary\">\n",
    "    <div class = \"panel-heading\"> \n",
    "    <span style=\"color:white\"><b>Extra Credit</b>\n",
    "    </div>\n",
    "    <div class = \"panel-body\"> \n",
    "    Create a neural network model with the following architecture and store this in a variable called <code>nnet_model_2</code>:\n",
    "    <ul>\n",
    "        <li>A normalization layer with means and standard deviations populated by the training data <code>X_train</code>.</li>\n",
    "        <li>A hidden layer with 128 nodes and ReLU activation function, and apply dropout with 10% probability.</li>\n",
    "        <li>A second hidden layer with 256 nodes and ReLU activation function, and apply dropout with 10% probability.</li>\n",
    "        <li>A third hidden layer with 256 nodes and ReLU activation function, and apply dropout with 10% probability.</li>\n",
    "        <li>An output layer with 1 node, since we are only interested in predicting the probability of loan defaults. We will not apply the sigmoid activation function yet, until after training is complete.</li>\n",
    "    </ul>\n",
    "    Configure the tensorflow model compilier for <code>nnet_model_2</code>, with the following specifications:\n",
    "    <ul>\n",
    "        <li><code>loss</code>: <code>tf.keras.losses.BinaryCrossentropy(from_logits = True)</code></li>\n",
    "        <li><code>optimizer</code>: Use the Adam Optimization Algorithm as the optimizer (tensorflow function is <code>ts.keras.optimizers.Adam()</code>. Set the learning rate equal to 0.001.</li>\n",
    "    </ul>\n",
    "    Train the neural network on the training data (<code>X_train</code> and <code>y_train</code>). Set the following additional parameters in the <code>.fit()</code> method:\n",
    "    <ul>\n",
    "        <li><code>validation_data = (X_test, y_test)</code>: this allows for the evaluation of model performance on the test data during training.</li>\n",
    "        <li><code>epochs = 20</code></li>\n",
    "        <li><code>shuffle = True</code></li>\n",
    "        <li><code>batch_size = 100</code></li>\n",
    "    </ul>\n",
    "    Plot both the training loss and test loss on a single line plot. Label the axes and provide a legend.\n",
    "    <br><br>\n",
    "    Use <code>nnet_model_2</code> to calculate the predicted probabilities of default for the test data (use a sigmoid activation function), and store those predictions in a variable called <code>probs_2</code>. Calculate the ROC curve and the AUC for <code>probs_2</code>, and plot the ROC curve on a line plot. Provide axis labels and a title for the plot.\n",
    "    <br><br>\n",
    "    Did <code>nnet_model_2</code> outperform <code>nnet_model_1</code>?\n",
    "    <p style=\"text-align:right;\">[20pts]</p>\n",
    "    </div>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}