Charity_Sucess_Deep_Learning.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Deep Learning Charity Success"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import our dependencies\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler,OneHotEncoder\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from sklearn.metrics import precision_score, recall_score, f1_score\n",
    "import numpy as np\n",
    "import os\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load/Evaluate charity data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "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>EIN</th>\n",
       "      <th>NAME</th>\n",
       "      <th>APPLICATION_TYPE</th>\n",
       "      <th>AFFILIATION</th>\n",
       "      <th>CLASSIFICATION</th>\n",
       "      <th>USE_CASE</th>\n",
       "      <th>ORGANIZATION</th>\n",
       "      <th>STATUS</th>\n",
       "      <th>INCOME_AMT</th>\n",
       "      <th>SPECIAL_CONSIDERATIONS</th>\n",
       "      <th>ASK_AMT</th>\n",
       "      <th>IS_SUCCESSFUL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>10520599</td>\n",
       "      <td>BLUE KNIGHTS MOTORCYCLE CLUB</td>\n",
       "      <td>T10</td>\n",
       "      <td>Independent</td>\n",
       "      <td>C1000</td>\n",
       "      <td>ProductDev</td>\n",
       "      <td>Association</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>5000</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>10531628</td>\n",
       "      <td>AMERICAN CHESAPEAKE CLUB CHARITABLE TR</td>\n",
       "      <td>T3</td>\n",
       "      <td>Independent</td>\n",
       "      <td>C2000</td>\n",
       "      <td>Preservation</td>\n",
       "      <td>Co-operative</td>\n",
       "      <td>1</td>\n",
       "      <td>1-9999</td>\n",
       "      <td>N</td>\n",
       "      <td>108590</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10547893</td>\n",
       "      <td>ST CLOUD PROFESSIONAL FIREFIGHTERS</td>\n",
       "      <td>T5</td>\n",
       "      <td>CompanySponsored</td>\n",
       "      <td>C3000</td>\n",
       "      <td>ProductDev</td>\n",
       "      <td>Association</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>5000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>10553066</td>\n",
       "      <td>SOUTHSIDE ATHLETIC ASSOCIATION</td>\n",
       "      <td>T3</td>\n",
       "      <td>CompanySponsored</td>\n",
       "      <td>C2000</td>\n",
       "      <td>Preservation</td>\n",
       "      <td>Trust</td>\n",
       "      <td>1</td>\n",
       "      <td>10000-24999</td>\n",
       "      <td>N</td>\n",
       "      <td>6692</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>10556103</td>\n",
       "      <td>GENETIC RESEARCH INSTITUTE OF THE DESERT</td>\n",
       "      <td>T3</td>\n",
       "      <td>Independent</td>\n",
       "      <td>C1000</td>\n",
       "      <td>Heathcare</td>\n",
       "      <td>Trust</td>\n",
       "      <td>1</td>\n",
       "      <td>100000-499999</td>\n",
       "      <td>N</td>\n",
       "      <td>142590</td>\n",
       "      <td>1</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34294</th>\n",
       "      <td>996009318</td>\n",
       "      <td>THE LIONS CLUB OF HONOLULU KAMEHAMEHA</td>\n",
       "      <td>T4</td>\n",
       "      <td>Independent</td>\n",
       "      <td>C1000</td>\n",
       "      <td>ProductDev</td>\n",
       "      <td>Association</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>5000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34295</th>\n",
       "      <td>996010315</td>\n",
       "      <td>INTERNATIONAL ASSOCIATION OF LIONS CLUBS</td>\n",
       "      <td>T4</td>\n",
       "      <td>CompanySponsored</td>\n",
       "      <td>C3000</td>\n",
       "      <td>ProductDev</td>\n",
       "      <td>Association</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>5000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34296</th>\n",
       "      <td>996012607</td>\n",
       "      <td>PTA HAWAII CONGRESS</td>\n",
       "      <td>T3</td>\n",
       "      <td>CompanySponsored</td>\n",
       "      <td>C2000</td>\n",
       "      <td>Preservation</td>\n",
       "      <td>Association</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>5000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34297</th>\n",
       "      <td>996015768</td>\n",
       "      <td>AMERICAN FEDERATION OF GOVERNMENT EMPLOYEES LO...</td>\n",
       "      <td>T5</td>\n",
       "      <td>Independent</td>\n",
       "      <td>C3000</td>\n",
       "      <td>ProductDev</td>\n",
       "      <td>Association</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>5000</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34298</th>\n",
       "      <td>996086871</td>\n",
       "      <td>WATERHOUSE CHARITABLE TR</td>\n",
       "      <td>T3</td>\n",
       "      <td>Independent</td>\n",
       "      <td>C1000</td>\n",
       "      <td>Preservation</td>\n",
       "      <td>Co-operative</td>\n",
       "      <td>1</td>\n",
       "      <td>1M-5M</td>\n",
       "      <td>N</td>\n",
       "      <td>36500179</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>34299 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             EIN                                               NAME  \\\n",
       "0       10520599                       BLUE KNIGHTS MOTORCYCLE CLUB   \n",
       "1       10531628             AMERICAN CHESAPEAKE CLUB CHARITABLE TR   \n",
       "2       10547893                 ST CLOUD PROFESSIONAL FIREFIGHTERS   \n",
       "3       10553066                     SOUTHSIDE ATHLETIC ASSOCIATION   \n",
       "4       10556103           GENETIC RESEARCH INSTITUTE OF THE DESERT   \n",
       "...          ...                                                ...   \n",
       "34294  996009318              THE LIONS CLUB OF HONOLULU KAMEHAMEHA   \n",
       "34295  996010315           INTERNATIONAL ASSOCIATION OF LIONS CLUBS   \n",
       "34296  996012607                                PTA HAWAII CONGRESS   \n",
       "34297  996015768  AMERICAN FEDERATION OF GOVERNMENT EMPLOYEES LO...   \n",
       "34298  996086871                           WATERHOUSE CHARITABLE TR   \n",
       "\n",
       "      APPLICATION_TYPE       AFFILIATION CLASSIFICATION      USE_CASE  \\\n",
       "0                  T10       Independent          C1000    ProductDev   \n",
       "1                   T3       Independent          C2000  Preservation   \n",
       "2                   T5  CompanySponsored          C3000    ProductDev   \n",
       "3                   T3  CompanySponsored          C2000  Preservation   \n",
       "4                   T3       Independent          C1000     Heathcare   \n",
       "...                ...               ...            ...           ...   \n",
       "34294               T4       Independent          C1000    ProductDev   \n",
       "34295               T4  CompanySponsored          C3000    ProductDev   \n",
       "34296               T3  CompanySponsored          C2000  Preservation   \n",
       "34297               T5       Independent          C3000    ProductDev   \n",
       "34298               T3       Independent          C1000  Preservation   \n",
       "\n",
       "       ORGANIZATION  STATUS     INCOME_AMT SPECIAL_CONSIDERATIONS   ASK_AMT  \\\n",
       "0       Association       1              0                      N      5000   \n",
       "1      Co-operative       1         1-9999                      N    108590   \n",
       "2       Association       1              0                      N      5000   \n",
       "3             Trust       1    10000-24999                      N      6692   \n",
       "4             Trust       1  100000-499999                      N    142590   \n",
       "...             ...     ...            ...                    ...       ...   \n",
       "34294   Association       1              0                      N      5000   \n",
       "34295   Association       1              0                      N      5000   \n",
       "34296   Association       1              0                      N      5000   \n",
       "34297   Association       1              0                      N      5000   \n",
       "34298  Co-operative       1          1M-5M                      N  36500179   \n",
       "\n",
       "       IS_SUCCESSFUL  \n",
       "0                  1  \n",
       "1                  1  \n",
       "2                  0  \n",
       "3                  1  \n",
       "4                  1  \n",
       "...              ...  \n",
       "34294              0  \n",
       "34295              0  \n",
       "34296              0  \n",
       "34297              1  \n",
       "34298              0  \n",
       "\n",
       "[34299 rows x 12 columns]"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#  Import and read the charity_data.csv.\n",
    "application_df = pd.read_csv(\"Resources/charity_data.csv\")\n",
    "application_df"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "the 'EIN' and 'NAME' columns contain unnecessary data for model training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "EIN                       34299\n",
       "NAME                      19568\n",
       "APPLICATION_TYPE             17\n",
       "AFFILIATION                   6\n",
       "CLASSIFICATION               71\n",
       "USE_CASE                      5\n",
       "ORGANIZATION                  4\n",
       "STATUS                        2\n",
       "INCOME_AMT                    9\n",
       "SPECIAL_CONSIDERATIONS        2\n",
       "ASK_AMT                    8747\n",
       "IS_SUCCESSFUL                 2\n",
       "dtype: int64"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Determine the number of unique values in each column.\n",
    "application_df.nunique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Drop the non-beneficial ID columns, 'EIN' and 'NAME'.\n",
    "application_df = application_df.drop(columns=[\"EIN\", \"NAME\"])"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "'Application_type\" and \"Classification\" could be binned to reduce model confusion.\n",
    "\"ask_amt\" will not be binned as its values will be a useful factor after standardization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "T3     27037\n",
       "T4      1542\n",
       "T6      1216\n",
       "T5      1173\n",
       "T19     1065\n",
       "T8       737\n",
       "T7       725\n",
       "T10      528\n",
       "T9       156\n",
       "T13       66\n",
       "T12       27\n",
       "T2        16\n",
       "T25        3\n",
       "T14        3\n",
       "T29        2\n",
       "T15        2\n",
       "T17        1\n",
       "Name: APPLICATION_TYPE, dtype: int64"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Look at APPLICATION_TYPE value counts for binning\n",
    "application_type_count = application_df.APPLICATION_TYPE.value_counts()\n",
    "application_type_count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "C1000    17326\n",
       "C2000     6074\n",
       "C1200     4837\n",
       "C3000     1918\n",
       "C2100     1883\n",
       "C7000      777\n",
       "C1700      287\n",
       "C4000      194\n",
       "C5000      116\n",
       "C1270      114\n",
       "C2700      104\n",
       "C2800       95\n",
       "C7100       75\n",
       "C1300       58\n",
       "C1280       50\n",
       "C1230       36\n",
       "C1400       34\n",
       "C7200       32\n",
       "C2300       32\n",
       "C1240       30\n",
       "C8000       20\n",
       "C7120       18\n",
       "C1500       16\n",
       "C1800       15\n",
       "C6000       15\n",
       "C1250       14\n",
       "C8200       11\n",
       "C1238       10\n",
       "C1278       10\n",
       "C1235        9\n",
       "C1237        9\n",
       "C7210        7\n",
       "C2400        6\n",
       "C1720        6\n",
       "C4100        6\n",
       "C1257        5\n",
       "C1600        5\n",
       "C1260        3\n",
       "C2710        3\n",
       "C0           3\n",
       "C3200        2\n",
       "C1234        2\n",
       "C1246        2\n",
       "C1267        2\n",
       "C1256        2\n",
       "C2190        1\n",
       "C4200        1\n",
       "C2600        1\n",
       "C5200        1\n",
       "C1370        1\n",
       "Name: CLASSIFICATION, dtype: int64"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Look at the Classification value counts for binning\n",
    "classification_count = application_df.CLASSIFICATION.value_counts()\n",
    "classification_count.head(50)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "T3 Application types, and C1000 classification types, heavily outnumber the other features. This may cause unwanted model performance later"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: ylabel='Density'>"
      ]
     },
     "execution_count": 67,
     "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": [
    "# Visualize the density of the APPLICATION_TYPE column\n",
    "application_type_count.plot.density()\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looking to find which value to best bin the APPLICATION_TYPE column, the plot \"falls off\" at an approximate count of 5-6000. As predicted the skewed quantity of T3 loans is showing signs of negatively influencing the model. If we were to bin our types at this level we would only have 3 buckets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: ylabel='Density'>"
      ]
     },
     "execution_count": 68,
     "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": [
    "# Vizualize the density without the T3 type\n",
    "\n",
    "reduced_application_type_count = application_type_count[~(application_type_count.index=='T3')]\n",
    "\n",
    "reduced_application_type_count.plot.density()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By removing the T3 loans we can see a better binning value would be at 1000."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: ylabel='Density'>"
      ]
     },
     "execution_count": 69,
     "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": [
    "# Visualize the density of the CLASSIFICATION column\n",
    "classification_count.plot.density()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The density of the Classification column is similar to that of the application_type, this may cause unwanted performance issues."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: ylabel='Density'>"
      ]
     },
     "execution_count": 70,
     "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": [
    "# Visualize the density without the C1000 type\n",
    "\n",
    "reduced_classification_count = classification_count[~(classification_count.index=='C1000')]\n",
    "reduced_classification_count.plot.density()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "While the change isnt drastic, it does yield a clearer value for binning, at around 500. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(T3       27037\n",
       " T4        1542\n",
       " T6        1216\n",
       " T5        1173\n",
       " T19       1065\n",
       " T8         737\n",
       " T7         725\n",
       " T10        528\n",
       " Other      276\n",
       " Name: APPLICATION_TYPE, dtype: int64,\n",
       " C1000    17326\n",
       " C2000     6074\n",
       " C1200     4837\n",
       " Other     2261\n",
       " C3000     1918\n",
       " C2100     1883\n",
       " Name: CLASSIFICATION, dtype: int64)"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Bin the APPLICATION_TYPE and CLASSIFICATION columns\n",
    "replace_application = list(application_type_count[application_type_count < 500].index)\n",
    "replace_classification = list(classification_count[classification_count < 1000].index)\n",
    "\n",
    "# Replace in dataframe\n",
    "for app in replace_application:\n",
    "    application_df.APPLICATION_TYPE = application_df.APPLICATION_TYPE.replace(app,\"Other\")\n",
    "\n",
    "for class_ in replace_classification:\n",
    "    application_df.CLASSIFICATION = application_df.CLASSIFICATION.replace(class_,\"Other\")\n",
    "\n",
    "# confirm binning\n",
    "application_df.APPLICATION_TYPE.value_counts(), application_df.CLASSIFICATION.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "7262"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "27037"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "34299"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "APPLICATION_TYPE             8\n",
       "AFFILIATION                  6\n",
       "CLASSIFICATION               5\n",
       "USE_CASE                     4\n",
       "ORGANIZATION                 4\n",
       "STATUS                       1\n",
       "INCOME_AMT                   9\n",
       "SPECIAL_CONSIDERATIONS       1\n",
       "ASK_AMT                   1365\n",
       "IS_SUCCESSFUL                2\n",
       "dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "APPLICATION_TYPE             1\n",
       "AFFILIATION                  6\n",
       "CLASSIFICATION               6\n",
       "USE_CASE                     5\n",
       "ORGANIZATION                 4\n",
       "STATUS                       2\n",
       "INCOME_AMT                   9\n",
       "SPECIAL_CONSIDERATIONS       2\n",
       "ASK_AMT                   7423\n",
       "IS_SUCCESSFUL                2\n",
       "dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "APPLICATION_TYPE             9\n",
       "AFFILIATION                  6\n",
       "CLASSIFICATION               6\n",
       "USE_CASE                     5\n",
       "ORGANIZATION                 4\n",
       "STATUS                       2\n",
       "INCOME_AMT                   9\n",
       "SPECIAL_CONSIDERATIONS       2\n",
       "ASK_AMT                   8747\n",
       "IS_SUCCESSFUL                2\n",
       "dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Evaluate the data with/without the T3 application type\n",
    "\n",
    "no_t3 = application_df[~(application_df.APPLICATION_TYPE=='T3')].copy()\n",
    "t3_only = application_df[application_df.APPLICATION_TYPE=='T3'].copy()\n",
    "\n",
    "display(len(no_t3),len(t3_only), len(application_df))\n",
    "display(no_t3.nunique(),t3_only.nunique(), application_df.nunique())"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By comparing the distribution of features between the no_t3 dataframe, and the t3_only dataframe, its apparent that dropping the t3 loan types from our analysis would negatively impact our models training.\n",
    "The t3_only dataframe is more representative of the original dataset than the no_t3 dataframe is. In order to correct this we will oversample the minority loan types in order to better represent them in our analysis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/homebrew/Caskroom/miniforge/base/lib/python3.10/site-packages/sklearn/preprocessing/_encoders.py:828: FutureWarning: `sparse` was renamed to `sparse_output` in version 1.2 and will be removed in 1.4. `sparse_output` is ignored unless you leave `sparse` to its default value.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "# Encode categorical variables using one-hot encoding\n",
    "application_cat = application_df.dtypes[application_df.dtypes == \"object\"].index.tolist()\n",
    "\n",
    "# Create a OneHotEncoder instance\n",
    "enc = OneHotEncoder(sparse=False)\n",
    "\n",
    "# Fit and transform the OneHotEncoder using the categorical variable list\n",
    "encode_df = pd.DataFrame(enc.fit_transform(application_df[application_cat]))\n",
    "\n",
    "\n",
    "# Add the encoded variable names to the dataframe\n",
    "encode_df.columns = enc.get_feature_names_out(application_cat)\n",
    "\n",
    "# Merge one-hot encoded features and drop the originals\n",
    "encoded_application_df = application_df.merge(encode_df,left_index=True,right_index=True).drop(columns=application_cat).copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "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>STATUS</th>\n",
       "      <th>ASK_AMT</th>\n",
       "      <th>IS_SUCCESSFUL</th>\n",
       "      <th>APPLICATION_TYPE_Other</th>\n",
       "      <th>APPLICATION_TYPE_T10</th>\n",
       "      <th>APPLICATION_TYPE_T19</th>\n",
       "      <th>APPLICATION_TYPE_T4</th>\n",
       "      <th>APPLICATION_TYPE_T5</th>\n",
       "      <th>APPLICATION_TYPE_T6</th>\n",
       "      <th>APPLICATION_TYPE_T7</th>\n",
       "      <th>...</th>\n",
       "      <th>INCOME_AMT_10000-24999</th>\n",
       "      <th>INCOME_AMT_100000-499999</th>\n",
       "      <th>INCOME_AMT_10M-50M</th>\n",
       "      <th>INCOME_AMT_1M-5M</th>\n",
       "      <th>INCOME_AMT_25000-99999</th>\n",
       "      <th>INCOME_AMT_50M+</th>\n",
       "      <th>INCOME_AMT_5M-10M</th>\n",
       "      <th>SPECIAL_CONSIDERATIONS_N</th>\n",
       "      <th>SPECIAL_CONSIDERATIONS_Y</th>\n",
       "      <th>APPLICATION_TYPE_T3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>5000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>108590</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>5000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>6692</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>142590</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</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",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54069</th>\n",
       "      <td>1</td>\n",
       "      <td>5000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54070</th>\n",
       "      <td>1</td>\n",
       "      <td>5000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54071</th>\n",
       "      <td>1</td>\n",
       "      <td>5000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54072</th>\n",
       "      <td>1</td>\n",
       "      <td>5000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54073</th>\n",
       "      <td>1</td>\n",
       "      <td>5000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>54074 rows × 44 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       STATUS  ASK_AMT  IS_SUCCESSFUL  APPLICATION_TYPE_Other  \\\n",
       "0           1     5000              1                     0.0   \n",
       "1           1   108590              1                     0.0   \n",
       "2           1     5000              0                     0.0   \n",
       "3           1     6692              1                     0.0   \n",
       "4           1   142590              1                     0.0   \n",
       "...       ...      ...            ...                     ...   \n",
       "54069       1     5000              1                     0.0   \n",
       "54070       1     5000              1                     0.0   \n",
       "54071       1     5000              0                     0.0   \n",
       "54072       1     5000              0                     0.0   \n",
       "54073       1     5000              1                     0.0   \n",
       "\n",
       "       APPLICATION_TYPE_T10  APPLICATION_TYPE_T19  APPLICATION_TYPE_T4  \\\n",
       "0                       1.0                   0.0                  0.0   \n",
       "1                       0.0                   0.0                  0.0   \n",
       "2                       0.0                   0.0                  0.0   \n",
       "3                       0.0                   0.0                  0.0   \n",
       "4                       0.0                   0.0                  0.0   \n",
       "...                     ...                   ...                  ...   \n",
       "54069                   0.0                   0.0                  0.0   \n",
       "54070                   0.0                   0.0                  0.0   \n",
       "54071                   0.0                   0.0                  1.0   \n",
       "54072                   0.0                   1.0                  0.0   \n",
       "54073                   1.0                   0.0                  0.0   \n",
       "\n",
       "       APPLICATION_TYPE_T5  APPLICATION_TYPE_T6  APPLICATION_TYPE_T7  ...  \\\n",
       "0                      0.0                  0.0                  0.0  ...   \n",
       "1                      0.0                  0.0                  0.0  ...   \n",
       "2                      1.0                  0.0                  0.0  ...   \n",
       "3                      0.0                  0.0                  0.0  ...   \n",
       "4                      0.0                  0.0                  0.0  ...   \n",
       "...                    ...                  ...                  ...  ...   \n",
       "54069                  0.0                  0.0                  1.0  ...   \n",
       "54070                  0.0                  0.0                  1.0  ...   \n",
       "54071                  0.0                  0.0                  0.0  ...   \n",
       "54072                  0.0                  0.0                  0.0  ...   \n",
       "54073                  0.0                  0.0                  0.0  ...   \n",
       "\n",
       "       INCOME_AMT_10000-24999  INCOME_AMT_100000-499999  INCOME_AMT_10M-50M  \\\n",
       "0                         0.0                       0.0                 0.0   \n",
       "1                         0.0                       0.0                 0.0   \n",
       "2                         0.0                       0.0                 0.0   \n",
       "3                         1.0                       0.0                 0.0   \n",
       "4                         0.0                       1.0                 0.0   \n",
       "...                       ...                       ...                 ...   \n",
       "54069                     0.0                       0.0                 0.0   \n",
       "54070                     0.0                       0.0                 0.0   \n",
       "54071                     0.0                       0.0                 0.0   \n",
       "54072                     0.0                       0.0                 0.0   \n",
       "54073                     0.0                       0.0                 0.0   \n",
       "\n",
       "       INCOME_AMT_1M-5M  INCOME_AMT_25000-99999  INCOME_AMT_50M+  \\\n",
       "0                   0.0                     0.0              0.0   \n",
       "1                   0.0                     0.0              0.0   \n",
       "2                   0.0                     0.0              0.0   \n",
       "3                   0.0                     0.0              0.0   \n",
       "4                   0.0                     0.0              0.0   \n",
       "...                 ...                     ...              ...   \n",
       "54069               0.0                     0.0              0.0   \n",
       "54070               0.0                     0.0              0.0   \n",
       "54071               0.0                     0.0              0.0   \n",
       "54072               0.0                     0.0              0.0   \n",
       "54073               0.0                     0.0              0.0   \n",
       "\n",
       "       INCOME_AMT_5M-10M  SPECIAL_CONSIDERATIONS_N  SPECIAL_CONSIDERATIONS_Y  \\\n",
       "0                    0.0                       1.0                       0.0   \n",
       "1                    0.0                       1.0                       0.0   \n",
       "2                    0.0                       1.0                       0.0   \n",
       "3                    0.0                       1.0                       0.0   \n",
       "4                    0.0                       1.0                       0.0   \n",
       "...                  ...                       ...                       ...   \n",
       "54069                0.0                       1.0                       0.0   \n",
       "54070                0.0                       1.0                       0.0   \n",
       "54071                0.0                       1.0                       0.0   \n",
       "54072                0.0                       1.0                       0.0   \n",
       "54073                0.0                       1.0                       0.0   \n",
       "\n",
       "       APPLICATION_TYPE_T3  \n",
       "0                      0.0  \n",
       "1                      1.0  \n",
       "2                      0.0  \n",
       "3                      1.0  \n",
       "4                      1.0  \n",
       "...                    ...  \n",
       "54069                  0.0  \n",
       "54070                  0.0  \n",
       "54071                  0.0  \n",
       "54072                  0.0  \n",
       "54073                  0.0  \n",
       "\n",
       "[54074 rows x 44 columns]"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Oversample the data using the SMOTE algorithm\n",
    "smote = SMOTE(random_state=1, sampling_strategy='minority')\n",
    "\n",
    "\n",
    "over_x = encoded_application_df.drop(columns=[\"APPLICATION_TYPE_T3\"])\n",
    "over_y = encoded_application_df[\"APPLICATION_TYPE_T3\"]\n",
    "\n",
    "# apply SMOTE to the data\n",
    "X_resampled, y_resampled = smote.fit_resample(over_x, over_y)\n",
    "\n",
    "# Convert X and y resampled to a DataFrame and concatenate\n",
    "X_resampled_df = pd.DataFrame(X_resampled, columns=over_x.columns)\n",
    "y_resampled_df = pd.DataFrame(y_resampled, columns=['APPLICATION_TYPE_T3'])\n",
    "enc_smoted_application_df = pd.concat([X_resampled_df,y_resampled_df],axis=1)\n",
    "\n",
    "# verify the resampled data\n",
    "enc_smoted_application_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split our preprocessed data into our features and target arrays\n",
    "\n",
    "y = enc_smoted_application_df[\"IS_SUCCESSFUL\"]\n",
    "X = enc_smoted_application_df.drop(columns=[\"IS_SUCCESSFUL\"])\n",
    "\n",
    "# Split the preprocessed data into a training and testing dataset\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=52)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Scale the data\n",
    "\n",
    "# Create a StandardScaler instances\n",
    "scaler = StandardScaler()\n",
    "\n",
    "# Fit the StandardScaler\n",
    "X_scaler = scaler.fit(X_train)\n",
    "\n",
    "# Scale the data\n",
    "X_train_scaled = X_scaler.transform(X_train)\n",
    "X_test_scaled = X_scaler.transform(X_test)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Compile, Train, and Evaluate the Model"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Model 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " dense_12 (Dense)            (None, 30)                1320      \n",
      "                                                                 \n",
      " dense_13 (Dense)            (None, 15)                465       \n",
      "                                                                 \n",
      " dense_14 (Dense)            (None, 8)                 128       \n",
      "                                                                 \n",
      " dense_15 (Dense)            (None, 1)                 9         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1,922\n",
      "Trainable params: 1,922\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Define the model input, hidden, and output layers\n",
    "number_input_features = len(X_train_scaled[0])\n",
    "hidden_nodes_layer1 = 30\n",
    "hidden_nodes_layer2 = 15\n",
    "hidden_nodes_layer3 = 8\n",
    "\n",
    "\n",
    "application_nn = tf.keras.models.Sequential()\n",
    "\n",
    "# First hidden layer\n",
    "application_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer1,\n",
    "        input_dim=number_input_features,\n",
    "        activation='relu'\n",
    "    )\n",
    ")\n",
    "\n",
    "# Second hidden layer\n",
    "application_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer2,\n",
    "        activation='selu'\n",
    "    )\n",
    ")\n",
    "\n",
    "# Third hidden layer\n",
    "application_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer3,\n",
    "        activation='swish'\n",
    "    )\n",
    ")\n",
    "\n",
    "# Output layer\n",
    "application_nn.add(tf.keras.layers.Dense(\n",
    "    units=1, \n",
    "    activation=\"sigmoid\"\n",
    "    )\n",
    ")\n",
    "# Check the structure of the model\n",
    "application_nn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the checkpoint path and filenames\n",
    "os.makedirs(\n",
    "    \"output/Checkpoints/initialModel_checkpoints/\",\n",
    "    exist_ok=True\n",
    "    )\n",
    "checkpoint_path = \"output/Checkpoints/initialModel_checkpoints/weights.{epoch:02d}.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:`period` argument is deprecated. Please use `save_freq` to specify the frequency in number of batches seen.\n"
     ]
    }
   ],
   "source": [
    "# Compile the model\n",
    "application_nn.compile(\n",
    "    loss='binary_crossentropy',\n",
    "    optimizer='adam',\n",
    "    metrics=['accuracy']\n",
    "    )\n",
    "\n",
    "# Create a callback that saves the model's weights every epoch\n",
    "cp_callback = ModelCheckpoint(\n",
    "    filepath=checkpoint_path, \n",
    "    verbose=1, \n",
    "    save_weights_only=True,\n",
    "    save_freq='epoch',\n",
    "    period=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/25\n",
      "1268/1268 [==============================] - 12s 9ms/step - loss: 0.5588 - accuracy: 0.7195\n",
      "Epoch 2/25\n",
      "1268/1268 [==============================] - 11s 8ms/step - loss: 0.5355 - accuracy: 0.7324\n",
      "Epoch 3/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5316 - accuracy: 0.7361\n",
      "Epoch 4/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5302 - accuracy: 0.7368\n",
      "Epoch 5/25\n",
      "1263/1268 [============================>.] - ETA: 0s - loss: 0.5284 - accuracy: 0.7369\n",
      "Epoch 5: saving model to output/Checkpoints/initialModel_checkpoints/weights.05.hdf5\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5284 - accuracy: 0.7369\n",
      "Epoch 6/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5275 - accuracy: 0.7378\n",
      "Epoch 7/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5265 - accuracy: 0.7396\n",
      "Epoch 8/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5257 - accuracy: 0.7412\n",
      "Epoch 9/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5252 - accuracy: 0.7390\n",
      "Epoch 10/25\n",
      "1263/1268 [============================>.] - ETA: 0s - loss: 0.5249 - accuracy: 0.7395\n",
      "Epoch 10: saving model to output/Checkpoints/initialModel_checkpoints/weights.10.hdf5\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5248 - accuracy: 0.7397\n",
      "Epoch 11/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5242 - accuracy: 0.7403\n",
      "Epoch 12/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5240 - accuracy: 0.7402\n",
      "Epoch 13/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5235 - accuracy: 0.7416\n",
      "Epoch 14/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5228 - accuracy: 0.7411\n",
      "Epoch 15/25\n",
      "1266/1268 [============================>.] - ETA: 0s - loss: 0.5230 - accuracy: 0.7406\n",
      "Epoch 15: saving model to output/Checkpoints/initialModel_checkpoints/weights.15.hdf5\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5230 - accuracy: 0.7407\n",
      "Epoch 16/25\n",
      "1268/1268 [==============================] - 12s 9ms/step - loss: 0.5222 - accuracy: 0.7410\n",
      "Epoch 17/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5219 - accuracy: 0.7428\n",
      "Epoch 18/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5213 - accuracy: 0.7421\n",
      "Epoch 19/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5214 - accuracy: 0.7415\n",
      "Epoch 20/25\n",
      "1268/1268 [==============================] - ETA: 0s - loss: 0.5209 - accuracy: 0.7427\n",
      "Epoch 20: saving model to output/Checkpoints/initialModel_checkpoints/weights.20.hdf5\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5209 - accuracy: 0.7427\n",
      "Epoch 21/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5203 - accuracy: 0.7445\n",
      "Epoch 22/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5203 - accuracy: 0.7434\n",
      "Epoch 23/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5199 - accuracy: 0.7450\n",
      "Epoch 24/25\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5200 - accuracy: 0.7439\n",
      "Epoch 25/25\n",
      "1263/1268 [============================>.] - ETA: 0s - loss: 0.5200 - accuracy: 0.7441\n",
      "Epoch 25: saving model to output/Checkpoints/initialModel_checkpoints/weights.25.hdf5\n",
      "1268/1268 [==============================] - 11s 9ms/step - loss: 0.5199 - accuracy: 0.7442\n"
     ]
    }
   ],
   "source": [
    "# Train the model\n",
    "fit_application_nn =\\\n",
    "    application_nn.fit(\n",
    "        X_train_scaled,\n",
    "        y_train,\n",
    "        epochs=25,\n",
    "        callbacks=[cp_callback]\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "423/423 - 2s - loss: 0.5320 - accuracy: 0.7350 - 2s/epoch - 4ms/step\n",
      "423/423 [==============================] - 1s 2ms/step\n",
      "Initial Model\n",
      "Loss: 0.5320261120796204, Accuracy: 0.7349656224250793\n",
      "Precision: 0.7253812354613596, Recall: 0.793805685193042, F1: 0.7580525356202309\n"
     ]
    }
   ],
   "source": [
    "# Evaluate the model using the test data\n",
    "model_loss, model_accuracy =\\\n",
    "    application_nn.evaluate(\n",
    "        X_test_scaled,\n",
    "        y_test,\n",
    "        verbose=2,\n",
    "    )\n",
    "\n",
    "# get precision recall and f1 scores\n",
    "y_pred = application_nn.predict(X_test_scaled)\n",
    "y_pred = np.where(y_pred >= 0.5, 1, 0)\n",
    "\n",
    "precision = precision_score(y_test, y_pred)\n",
    "recall = recall_score(y_test, y_pred)\n",
    "f1 = f1_score(y_test, y_pred)\n",
    "\n",
    "\n",
    "print(\"Initial Model\")\n",
    "print(f\"Loss: {model_loss}, Accuracy: {model_accuracy}\")\n",
    "print(f\"Precision: {precision}, Recall: {recall}, F1: {f1}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Export our model to HDF5 file\n",
    "tf.keras.models.save_model(\n",
    "    application_nn,\n",
    "    \"output/Models/initialModel.h5\"\n",
    "    )"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Model 1 Summary\n",
    "This model fails to achieve the base target of the analysis, 75% accuracy. This model when ran through several iterations with various node and layer counts was limited to an accuracy score of 74%.\n",
    "\n",
    "As our data was oversampled to better accommodate one type of loan, and our models performance was lacking, one conclusion that can be drawn is that the \"T3\" \"APPLICATION_TYPE\" data was overfitting our initial model."
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Models 2/3\n",
    "\n",
    "The errors in the last model may have been influenced by the application type, these models train independently around that feature and test results on its counterpart dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Relaod the data\n",
    "application_df2 = pd.read_csv(\n",
    "    \"Resources/charity_data.csv\"\n",
    "    )\n",
    "\n",
    "# Drop the non-beneficial ID columns, 'EIN' and 'NAME'.\n",
    "application_df2 = application_df2.drop(\n",
    "    columns=[\"EIN\",\"NAME\"]\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(T3       27037\n",
       " T4        1542\n",
       " T6        1216\n",
       " T5        1173\n",
       " T19       1065\n",
       " T8         737\n",
       " T7         725\n",
       " T10        528\n",
       " Other      276\n",
       " Name: APPLICATION_TYPE, dtype: int64,\n",
       " C1000    17326\n",
       " C2000     6074\n",
       " C1200     4837\n",
       " Other     2261\n",
       " C3000     1918\n",
       " C2100     1883\n",
       " Name: CLASSIFICATION, dtype: int64)"
      ]
     },
     "execution_count": 84,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Gather at APPLICATION_TYPE and CLASSIFICATION value counts for binning\n",
    "application_type_count = application_df2.APPLICATION_TYPE.value_counts()\n",
    "classification_count = application_df2.CLASSIFICATION.value_counts()\n",
    "\n",
    "# Bin the APPLICATION_TYPE and CLASSIFICATION columns\n",
    "replace_application = list(application_type_count[application_type_count < 500].index)\n",
    "replace_classification = list(classification_count[classification_count < 1000].index)\n",
    "\n",
    "# Replace in dataframe\n",
    "for app in replace_application:\n",
    "    application_df2.APPLICATION_TYPE = application_df2.APPLICATION_TYPE.replace(app,\"Other\")\n",
    "\n",
    "for class_ in replace_classification:\n",
    "    application_df2.CLASSIFICATION = application_df2.CLASSIFICATION.replace(class_,\"Other\")\n",
    "\n",
    "# confirm binning\n",
    "application_df2.APPLICATION_TYPE.value_counts(), application_df2.CLASSIFICATION.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/homebrew/Caskroom/miniforge/base/lib/python3.10/site-packages/sklearn/preprocessing/_encoders.py:828: FutureWarning: `sparse` was renamed to `sparse_output` in version 1.2 and will be removed in 1.4. `sparse_output` is ignored unless you leave `sparse` to its default value.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "# Encode application_df2\n",
    "\n",
    "# Encode categorical variables using one-hot encoding\n",
    "application_cat2 = application_df2.dtypes[application_df2.dtypes == \"object\"].index.tolist()\n",
    "# Create a OneHotEncoder instance\n",
    "enc2 = OneHotEncoder(sparse=False)\n",
    "# Fit and transform the OneHotEncoder using the categorical variable list\n",
    "encode_df2 = pd.DataFrame(enc2.fit_transform(application_df2[application_cat2]))\n",
    "# Add the encoded variable names to the dataframe\n",
    "encode_df2.columns = enc2.get_feature_names_out(application_cat2)\n",
    "# Merge one-hot encoded features and drop the originals\n",
    "encoded_application_df2 = application_df2.merge(encode_df2,left_index=True,right_index=True).drop(columns=application_cat2).copy()\n",
    "\n",
    "# split the data by t3 and no_t3\n",
    "encoded_no_t3 = encoded_application_df2[~(encoded_application_df2.APPLICATION_TYPE_T3==1)].copy()\n",
    "encoded_t3_only = encoded_application_df2[encoded_application_df2.APPLICATION_TYPE_T3==1].copy()\n",
    "\n",
    "# Split data into training and testing datasets\n",
    "y_no_t3 = encoded_no_t3[\"IS_SUCCESSFUL\"]\n",
    "X_no_t3 = encoded_no_t3.drop(columns=[\"IS_SUCCESSFUL\"])\n",
    "y_t3_only = encoded_t3_only[\"IS_SUCCESSFUL\"]\n",
    "X_t3_only = encoded_t3_only.drop(columns=[\"IS_SUCCESSFUL\"])\n",
    "\n",
    "# split the data into training and testing sets\n",
    "X_train_no_t3, X_test_no_t3, y_train_no_t3, y_test_no_t3 = train_test_split(X_no_t3, y_no_t3, random_state=52)\n",
    "X_train_t3_only, X_test_t3_only, y_train_t3_only, y_test_t3_only = train_test_split(X_t3_only, y_t3_only, random_state=52)\n",
    "\n",
    "\n",
    "# Scale the data\n",
    "scaler_no_t3 = StandardScaler()\n",
    "scaler_t3_only = StandardScaler()\n",
    "X_scaler_no_t3 = scaler_no_t3.fit(X_train_no_t3)\n",
    "X_scaler_t3_only = scaler_t3_only.fit(X_train_t3_only)\n",
    "\n",
    "X_train_no_t3_scaled = X_scaler_no_t3.transform(X_train_no_t3)\n",
    "X_test_no_t3_scaled = X_scaler_no_t3.transform(X_test_no_t3)\n",
    "X_train_t3_only_scaled = X_scaler_t3_only.transform(X_train_t3_only)\n",
    "X_test_t3_only_scaled = X_scaler_t3_only.transform(X_test_t3_only)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_4\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " dense_16 (Dense)            (None, 30)                1320      \n",
      "                                                                 \n",
      " dense_17 (Dense)            (None, 15)                465       \n",
      "                                                                 \n",
      " dense_18 (Dense)            (None, 1)                 16        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1,801\n",
      "Trainable params: 1,801\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# create the t3_only model\n",
    "\n",
    "# Define the model input, hidden, and output layers\n",
    "number_input_features = len(X_train_t3_only_scaled[0])\n",
    "hidden_nodes_layer1 = 30\n",
    "hidden_nodes_layer2 = 15\n",
    "\n",
    "t3_only_nn = tf.keras.models.Sequential()\n",
    "\n",
    "# First hidden layer\n",
    "t3_only_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer1,\n",
    "        input_dim=number_input_features,\n",
    "        activation='relu'\n",
    "    ))\n",
    "# Second hidden layer\n",
    "t3_only_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer2,\n",
    "        activation='relu'\n",
    "    ))\n",
    "# Output layer\n",
    "t3_only_nn.add(tf.keras.layers.Dense(\n",
    "    units=1, \n",
    "    activation=\"sigmoid\"\n",
    "    ))\n",
    "\n",
    "# Check the structure of the model\n",
    "t3_only_nn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_5\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " dense_19 (Dense)            (None, 30)                1320      \n",
      "                                                                 \n",
      " dense_20 (Dense)            (None, 15)                465       \n",
      "                                                                 \n",
      " dense_21 (Dense)            (None, 1)                 16        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1,801\n",
      "Trainable params: 1,801\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# create the no_t3 model\n",
    "\n",
    "# Define the model input, hidden, and output layers\n",
    "number_input_features = len(X_train_no_t3_scaled[0])\n",
    "hidden_nodes_layer1 = 30\n",
    "hidden_nodes_layer2 = 15\n",
    "\n",
    "no_t3_nn = tf.keras.models.Sequential()\n",
    "\n",
    "# First hidden layer\n",
    "no_t3_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer1,\n",
    "        input_dim=number_input_features,\n",
    "        activation='relu'\n",
    "    ))\n",
    "# Second hidden layer\n",
    "no_t3_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer2,\n",
    "        activation='relu'\n",
    "    ))\n",
    "# Output layer\n",
    "no_t3_nn.add(tf.keras.layers.Dense(\n",
    "    units=1, \n",
    "    activation=\"sigmoid\"\n",
    "    ))\n",
    "\n",
    "# Check the structure of the model\n",
    "no_t3_nn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the checkpoint path and filenames\n",
    "os.makedirs(\n",
    "    \"output/Checkpoints/no_t3_checkpoints/\",\n",
    "    exist_ok=True\n",
    "    )\n",
    "os.makedirs(\n",
    "    \"output/Checkpoints/t3_only_checkpoints/\",\n",
    "    exist_ok=True\n",
    "    )\n",
    "checkpoint_path_t3_only = \"output/Checkpoints/t3_only_checkpoints/weights.{epoch:02d}.hdf5\"\n",
    "checkpoint_path_no_t3 = \"output/Checkpoints/no_t3_checkpoints/weights.{epoch:02d}.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:`period` argument is deprecated. Please use `save_freq` to specify the frequency in number of batches seen.\n",
      "Epoch 1/25\n",
      "171/171 [==============================] - 2s 7ms/step - loss: 0.6021 - accuracy: 0.6992\n",
      "Epoch 2/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.5240 - accuracy: 0.7505\n",
      "Epoch 3/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.5095 - accuracy: 0.7510\n",
      "Epoch 4/25\n",
      "171/171 [==============================] - 1s 7ms/step - loss: 0.5024 - accuracy: 0.7584\n",
      "Epoch 5/25\n",
      "165/171 [===========================>..] - ETA: 0s - loss: 0.4989 - accuracy: 0.7604\n",
      "Epoch 5: saving model to output/Checkpoints/no_t3_checkpoints/weights.05.hdf5\n",
      "171/171 [==============================] - 1s 7ms/step - loss: 0.5001 - accuracy: 0.7589\n",
      "Epoch 6/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4973 - accuracy: 0.7600\n",
      "Epoch 7/25\n",
      "171/171 [==============================] - 1s 7ms/step - loss: 0.4953 - accuracy: 0.7633\n",
      "Epoch 8/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4936 - accuracy: 0.7607\n",
      "Epoch 9/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4922 - accuracy: 0.7600\n",
      "Epoch 10/25\n",
      "166/171 [============================>.] - ETA: 0s - loss: 0.4912 - accuracy: 0.7619\n",
      "Epoch 10: saving model to output/Checkpoints/no_t3_checkpoints/weights.10.hdf5\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4917 - accuracy: 0.7620\n",
      "Epoch 11/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4908 - accuracy: 0.7622\n",
      "Epoch 12/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4903 - accuracy: 0.7620\n",
      "Epoch 13/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4906 - accuracy: 0.7618\n",
      "Epoch 14/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4879 - accuracy: 0.7648\n",
      "Epoch 15/25\n",
      "166/171 [============================>.] - ETA: 0s - loss: 0.4864 - accuracy: 0.7649\n",
      "Epoch 15: saving model to output/Checkpoints/no_t3_checkpoints/weights.15.hdf5\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4871 - accuracy: 0.7640\n",
      "Epoch 16/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4875 - accuracy: 0.7629\n",
      "Epoch 17/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4863 - accuracy: 0.7626\n",
      "Epoch 18/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4860 - accuracy: 0.7620\n",
      "Epoch 19/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4864 - accuracy: 0.7653\n",
      "Epoch 20/25\n",
      "163/171 [===========================>..] - ETA: 0s - loss: 0.4850 - accuracy: 0.7694\n",
      "Epoch 20: saving model to output/Checkpoints/no_t3_checkpoints/weights.20.hdf5\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4857 - accuracy: 0.7686\n",
      "Epoch 21/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4842 - accuracy: 0.7622\n",
      "Epoch 22/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4840 - accuracy: 0.7705\n",
      "Epoch 23/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4839 - accuracy: 0.7668\n",
      "Epoch 24/25\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4831 - accuracy: 0.7639\n",
      "Epoch 25/25\n",
      "163/171 [===========================>..] - ETA: 0s - loss: 0.4809 - accuracy: 0.7692\n",
      "Epoch 25: saving model to output/Checkpoints/no_t3_checkpoints/weights.25.hdf5\n",
      "171/171 [==============================] - 1s 6ms/step - loss: 0.4829 - accuracy: 0.7677\n"
     ]
    }
   ],
   "source": [
    "# Compile the model\n",
    "no_t3_nn.compile(\n",
    "    loss='binary_crossentropy',\n",
    "    optimizer='adam',\n",
    "    metrics=['accuracy']\n",
    "    )\n",
    "\n",
    "# Create a callback that saves the model's weights every epoch\n",
    "cp_callback_no_t3 = ModelCheckpoint(\n",
    "    filepath=checkpoint_path_no_t3, \n",
    "    verbose=1, \n",
    "    save_weights_only=True,\n",
    "    save_freq='epoch',\n",
    "    period=5)\n",
    "\n",
    "# Train the model\n",
    "fit_no_t3_nn =\\\n",
    "    no_t3_nn.fit(\n",
    "        X_train_no_t3_scaled,\n",
    "        y_train_no_t3,\n",
    "        epochs=25,\n",
    "        callbacks=[cp_callback_no_t3]\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:`period` argument is deprecated. Please use `save_freq` to specify the frequency in number of batches seen.\n",
      "Epoch 1/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5927 - accuracy: 0.7048\n",
      "Epoch 2/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5696 - accuracy: 0.7206\n",
      "Epoch 3/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5669 - accuracy: 0.7220\n",
      "Epoch 4/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5662 - accuracy: 0.7215\n",
      "Epoch 5/25\n",
      "633/634 [============================>.] - ETA: 0s - loss: 0.5649 - accuracy: 0.7245\n",
      "Epoch 5: saving model to output/Checkpoints/t3_only_checkpoints/weights.05.hdf5\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5650 - accuracy: 0.7243\n",
      "Epoch 6/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5642 - accuracy: 0.7240\n",
      "Epoch 7/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5635 - accuracy: 0.7244\n",
      "Epoch 8/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5634 - accuracy: 0.7244\n",
      "Epoch 9/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5628 - accuracy: 0.7252\n",
      "Epoch 10/25\n",
      "632/634 [============================>.] - ETA: 0s - loss: 0.5626 - accuracy: 0.7255\n",
      "Epoch 10: saving model to output/Checkpoints/t3_only_checkpoints/weights.10.hdf5\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5626 - accuracy: 0.7255\n",
      "Epoch 11/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5624 - accuracy: 0.7251\n",
      "Epoch 12/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5618 - accuracy: 0.7251\n",
      "Epoch 13/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5620 - accuracy: 0.7253\n",
      "Epoch 14/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5621 - accuracy: 0.7250\n",
      "Epoch 15/25\n",
      "626/634 [============================>.] - ETA: 0s - loss: 0.5616 - accuracy: 0.7260\n",
      "Epoch 15: saving model to output/Checkpoints/t3_only_checkpoints/weights.15.hdf5\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5616 - accuracy: 0.7257\n",
      "Epoch 16/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5617 - accuracy: 0.7248\n",
      "Epoch 17/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5618 - accuracy: 0.7251\n",
      "Epoch 18/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5617 - accuracy: 0.7255\n",
      "Epoch 19/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5613 - accuracy: 0.7254\n",
      "Epoch 20/25\n",
      "634/634 [==============================] - ETA: 0s - loss: 0.5609 - accuracy: 0.7262\n",
      "Epoch 20: saving model to output/Checkpoints/t3_only_checkpoints/weights.20.hdf5\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5609 - accuracy: 0.7262\n",
      "Epoch 21/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5609 - accuracy: 0.7262\n",
      "Epoch 22/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5608 - accuracy: 0.7258\n",
      "Epoch 23/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5604 - accuracy: 0.7248\n",
      "Epoch 24/25\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5606 - accuracy: 0.7260\n",
      "Epoch 25/25\n",
      "629/634 [============================>.] - ETA: 0s - loss: 0.5595 - accuracy: 0.7270\n",
      "Epoch 25: saving model to output/Checkpoints/t3_only_checkpoints/weights.25.hdf5\n",
      "634/634 [==============================] - 4s 6ms/step - loss: 0.5602 - accuracy: 0.7263\n"
     ]
    }
   ],
   "source": [
    "# Compile the model\n",
    "t3_only_nn.compile(\n",
    "    loss='binary_crossentropy',\n",
    "    optimizer='adam',\n",
    "    metrics=['accuracy']\n",
    "    )\n",
    "\n",
    "# Create a callback that saves the model's weights every epoch\n",
    "cp_callback_t3_only = ModelCheckpoint(\n",
    "    filepath=checkpoint_path_t3_only, \n",
    "    verbose=1, \n",
    "    save_weights_only=True,\n",
    "    save_freq='epoch',\n",
    "    period=5)\n",
    "\n",
    "# Train the model\n",
    "fit_t3_only_nn =\\\n",
    "    t3_only_nn.fit(\n",
    "        X_train_t3_only_scaled,\n",
    "        y_train_t3_only,\n",
    "        epochs=25,\n",
    "        callbacks=[cp_callback_t3_only]\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "57/57 - 0s - loss: 0.4615 - accuracy: 0.7792 - 335ms/epoch - 6ms/step\n",
      "212/212 - 1s - loss: 0.5685 - accuracy: 0.7206 - 800ms/epoch - 4ms/step\n",
      "212/212 [==============================] - 0s 2ms/step\n",
      "57/57 [==============================] - 0s 2ms/step\n",
      "NO_T3 MODEL\n",
      "Loss: 0.461511492729187, Accuracy: 0.7791849970817566\n",
      "Precision: 0.5991514480723114, Recall: 0.8947658402203856, F1-Score: 0.7177107501933488\n",
      "T3_ONLY MODEL\n",
      "Loss: 0.5684725046157837, Accuracy: 0.7205621004104614\n",
      "Precision: 0.6496551724137931, Recall: 0.4931937172774869, F1-Score: 0.5607142857142857\n"
     ]
    }
   ],
   "source": [
    "# Evaluate the model using the test data\n",
    "model_loss_no_t3, model_accuracy_no_t3 =\\\n",
    "    no_t3_nn.evaluate(\n",
    "        X_test_no_t3_scaled,\n",
    "        y_test_no_t3,\n",
    "        verbose=2,\n",
    "    )\n",
    "model_loss_t3_only, model_accuracy_t3_only =\\\n",
    "    t3_only_nn.evaluate(\n",
    "        X_test_t3_only_scaled,\n",
    "        y_test_t3_only,\n",
    "        verbose=2,\n",
    "    )\n",
    "\n",
    "# make predictions on the opposite models test data\n",
    "predictions_no_t3 = no_t3_nn.predict(X_test_t3_only_scaled)\n",
    "predictions_t3_only = t3_only_nn.predict(X_test_no_t3_scaled)\n",
    "predictions_no_t3 = np.where(predictions_no_t3 >= 0.5, 1, 0)\n",
    "predictions_t3_only = np.where(predictions_t3_only >= 0.5, 1, 0)\n",
    "\n",
    "# get the precision, recall, f1-score, and support for both models\n",
    "\n",
    "# no_t3 model\n",
    "precision_no_t3 = precision_score(y_test_t3_only, predictions_no_t3)\n",
    "recall_no_t3 = recall_score(y_test_t3_only, predictions_no_t3)\n",
    "f1_no_t3 = f1_score(y_test_t3_only, predictions_no_t3)\n",
    "\n",
    "# t3_only model\n",
    "precision_t3_only = precision_score(y_test_no_t3,predictions_t3_only)\n",
    "recall_t3_only = recall_score(y_test_no_t3, predictions_t3_only)\n",
    "f1_t3_only = f1_score(y_test_no_t3, predictions_t3_only)\n",
    "\n",
    "print('\\n')\n",
    "print(\"NO_T3 MODEL\")\n",
    "print(f\"Loss: {model_loss_no_t3}, Accuracy: {model_accuracy_no_t3}\")\n",
    "print(f\"Precision: {precision_no_t3}, Recall: {recall_no_t3}, F1-Score: {f1_no_t3}\")\n",
    "print('\\n')\n",
    "print(\"T3_ONLY MODEL\")\n",
    "print(f\"Loss: {model_loss_t3_only}, Accuracy: {model_accuracy_t3_only}\")\n",
    "print(f\"Precision: {precision_t3_only}, Recall: {recall_t3_only}, F1-Score: {f1_t3_only}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Export our model to HDF5 file\n",
    "tf.keras.models.save_model(\n",
    "    no_t3_nn,\n",
    "    \"output/Models/no_t3.h5\"\n",
    "    )\n",
    "tf.keras.models.save_model(\n",
    "    t3_only_nn,\n",
    "    \"output/Models/t3_only.h5\"\n",
    "    )"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Model 2/3 Summary\n",
    "\n",
    "Unfortunately even separating the data by the overpopulated application_type didnt provide a successful outcome.\n",
    "The model trained on the no_t3 data performed better overall which came as a surprise as the t3_only dataset, while being limited in that one feature, was a better representation of the dataset as a whole. "
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Model 4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [],
   "source": [
    "# reload the data\n",
    "\n",
    "application_df3 = pd.read_csv(\n",
    "    \"Resources/charity_data.csv\")\n",
    "\n",
    "# drop the EIN and NAME columns\n",
    "application_df3 = application_df3.drop(\n",
    "    columns=[\"EIN\", \"NAME\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(T3       27037\n",
       " T4        1542\n",
       " T6        1216\n",
       " T5        1173\n",
       " T19       1065\n",
       " T8         737\n",
       " T7         725\n",
       " T10        528\n",
       " Other      276\n",
       " Name: APPLICATION_TYPE, dtype: int64,\n",
       " C1000    17326\n",
       " C2000     6074\n",
       " C1200     4837\n",
       " Other     2261\n",
       " C3000     1918\n",
       " C2100     1883\n",
       " Name: CLASSIFICATION, dtype: int64)"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Gather at APPLICATION_TYPE and CLASSIFICATION value counts for binning\n",
    "application_type_count3 = application_df3.APPLICATION_TYPE.value_counts()\n",
    "classification_count3 = application_df3.CLASSIFICATION.value_counts()\n",
    "\n",
    "# Bin the APPLICATION_TYPE and CLASSIFICATION columns\n",
    "replace_application3 = list(application_type_count3[application_type_count3 < 500].index)\n",
    "replace_classification3 = list(classification_count3[classification_count3 < 1000].index)\n",
    "\n",
    "# Replace in dataframe\n",
    "for app in replace_application3:\n",
    "    application_df3.APPLICATION_TYPE = application_df3.APPLICATION_TYPE.replace(app,\"Other\")\n",
    "\n",
    "for class_ in replace_classification3:\n",
    "    application_df3.CLASSIFICATION = application_df3.CLASSIFICATION.replace(class_,\"Other\")\n",
    "\n",
    "# confirm binning\n",
    "application_df3.APPLICATION_TYPE.value_counts(), application_df3.CLASSIFICATION.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/homebrew/Caskroom/miniforge/base/lib/python3.10/site-packages/sklearn/preprocessing/_encoders.py:828: FutureWarning: `sparse` was renamed to `sparse_output` in version 1.2 and will be removed in 1.4. `sparse_output` is ignored unless you leave `sparse` to its default value.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "# Encode categorical variables using one-hot encoding\n",
    "application_cat = application_df3.dtypes[application_df3.dtypes == \"object\"].index.tolist()\n",
    "\n",
    "# Create a OneHotEncoder instance\n",
    "enc = OneHotEncoder(sparse=False)\n",
    "\n",
    "# Fit and transform the OneHotEncoder using the categorical variable list\n",
    "encode_df = pd.DataFrame(enc.fit_transform(application_df3[application_cat]))\n",
    "\n",
    "\n",
    "# Add the encoded variable names to the dataframe\n",
    "encode_df.columns = enc.get_feature_names_out(application_cat)\n",
    "\n",
    "# Merge one-hot encoded features and drop the originals\n",
    "encoded_application_df3 = application_df3.merge(encode_df,left_index=True,right_index=True).drop(columns=application_cat).copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/homebrew/Caskroom/miniforge/base/lib/python3.10/site-packages/sklearn/cluster/_kmeans.py:870: FutureWarning: The default value of `n_init` will change from 10 to 'auto' in 1.4. Set the value of `n_init` explicitly to suppress the warning\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "from imblearn.under_sampling import ClusterCentroids\n",
    "\n",
    "# undersample the data\n",
    "under_x = encoded_application_df3.drop(columns=[\"APPLICATION_TYPE_T3\"])\n",
    "under_y = encoded_application_df3[\"APPLICATION_TYPE_T3\"]\n",
    "\n",
    "# apply the cc method\n",
    "cc = ClusterCentroids(random_state=1)\n",
    "under_x_resampled, under_y_resampled = cc.fit_resample(under_x, under_y)\n",
    "\n",
    "# Convert X and y resampled to a DataFrame and concatenate\n",
    "X_resampled_df = pd.DataFrame(under_x_resampled, columns=under_x.columns)\n",
    "y_resampled_df = pd.DataFrame(under_y_resampled, columns=['APPLICATION_TYPE_T3'])\n",
    "enc_cc_application_df = pd.concat([X_resampled_df,y_resampled_df],axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split our preprocessed data into our features and target arrays\n",
    "\n",
    "y = enc_cc_application_df[\"IS_SUCCESSFUL\"]\n",
    "X = enc_cc_application_df.drop(columns=[\"IS_SUCCESSFUL\"])\n",
    "\n",
    "# Split the preprocessed data into a training and testing dataset\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=52)\n",
    "\n",
    "# Scale the data\n",
    "\n",
    "# Create a StandardScaler instances\n",
    "scaler = StandardScaler()\n",
    "\n",
    "# Fit the StandardScaler\n",
    "X_scaler = scaler.fit(X_train)\n",
    "\n",
    "# Scale the data\n",
    "X_train_scaled = X_scaler.transform(X_train)\n",
    "X_test_scaled = X_scaler.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_6\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " dense_22 (Dense)            (None, 30)                1320      \n",
      "                                                                 \n",
      " dense_23 (Dense)            (None, 15)                465       \n",
      "                                                                 \n",
      " dense_24 (Dense)            (None, 8)                 128       \n",
      "                                                                 \n",
      " dense_25 (Dense)            (None, 1)                 9         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1,922\n",
      "Trainable params: 1,922\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Define the model input, hidden, and output layers\n",
    "number_input_features = len(X_train_scaled[0])\n",
    "hidden_nodes_layer1 = 30\n",
    "hidden_nodes_layer2 = 15\n",
    "hidden_nodes_layer3 = 8\n",
    "\n",
    "\n",
    "under_nn = tf.keras.models.Sequential()\n",
    "\n",
    "# First hidden layer\n",
    "under_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer1,\n",
    "        input_dim=number_input_features,\n",
    "        activation='relu'\n",
    "    )\n",
    ")\n",
    "\n",
    "# Second hidden layer\n",
    "under_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer2,\n",
    "        activation='selu'\n",
    "    )\n",
    ")\n",
    "\n",
    "# Third hidden layer\n",
    "under_nn.add(\n",
    "    tf.keras.layers.Dense(\n",
    "        units=hidden_nodes_layer3,\n",
    "        activation='swish'\n",
    "    )\n",
    ")\n",
    "\n",
    "# Output layer\n",
    "under_nn.add(tf.keras.layers.Dense(\n",
    "    units=1, \n",
    "    activation=\"sigmoid\"\n",
    "    )\n",
    ")\n",
    "# Check the structure of the model\n",
    "under_nn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the checkpoint path and filenames\n",
    "os.makedirs(\n",
    "    \"output/Checkpoints/undersampleModel_checkpoints/\",\n",
    "    exist_ok=True\n",
    "    )\n",
    "checkpoint_path_under = \"output/Checkpoints/undersampleModel/weights.{epoch:02d}.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:`period` argument is deprecated. Please use `save_freq` to specify the frequency in number of batches seen.\n"
     ]
    }
   ],
   "source": [
    "# Compile the model\n",
    "under_nn.compile(\n",
    "    loss='binary_crossentropy',\n",
    "    optimizer='adam',\n",
    "    metrics=['accuracy']\n",
    "    )\n",
    "\n",
    "# Create a callback that saves the model's weights every epoch\n",
    "cp_callback_under = ModelCheckpoint(\n",
    "    filepath=checkpoint_path_under, \n",
    "    verbose=1, \n",
    "    save_weights_only=True,\n",
    "    save_freq='epoch',\n",
    "    period=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/25\n",
      "341/341 [==============================] - 5s 13ms/step - loss: 0.5955 - accuracy: 0.6892\n",
      "Epoch 2/25\n",
      "341/341 [==============================] - 3s 9ms/step - loss: 0.5635 - accuracy: 0.7113\n",
      "Epoch 3/25\n",
      "341/341 [==============================] - 3s 9ms/step - loss: 0.5559 - accuracy: 0.7142\n",
      "Epoch 4/25\n",
      "341/341 [==============================] - 4s 10ms/step - loss: 0.5520 - accuracy: 0.7164\n",
      "Epoch 5/25\n",
      "341/341 [==============================] - ETA: 0s - loss: 0.5501 - accuracy: 0.7172\n",
      "Epoch 5: saving model to output/Checkpoints/undersampleModel/weights.05.hdf5\n",
      "341/341 [==============================] - 4s 11ms/step - loss: 0.5501 - accuracy: 0.7172\n",
      "Epoch 6/25\n",
      "341/341 [==============================] - 3s 9ms/step - loss: 0.5471 - accuracy: 0.7183\n",
      "Epoch 7/25\n",
      "341/341 [==============================] - 3s 10ms/step - loss: 0.5463 - accuracy: 0.7178\n",
      "Epoch 8/25\n",
      "341/341 [==============================] - 3s 9ms/step - loss: 0.5450 - accuracy: 0.7197\n",
      "Epoch 9/25\n",
      "341/341 [==============================] - 3s 9ms/step - loss: 0.5434 - accuracy: 0.7207\n",
      "Epoch 10/25\n",
      "338/341 [============================>.] - ETA: 0s - loss: 0.5428 - accuracy: 0.7209\n",
      "Epoch 10: saving model to output/Checkpoints/undersampleModel/weights.10.hdf5\n",
      "341/341 [==============================] - 3s 10ms/step - loss: 0.5426 - accuracy: 0.7208\n",
      "Epoch 11/25\n",
      "341/341 [==============================] - 3s 10ms/step - loss: 0.5415 - accuracy: 0.7235\n",
      "Epoch 12/25\n",
      "341/341 [==============================] - 3s 10ms/step - loss: 0.5402 - accuracy: 0.7228\n",
      "Epoch 13/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5398 - accuracy: 0.7253\n",
      "Epoch 14/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5392 - accuracy: 0.7229\n",
      "Epoch 15/25\n",
      "336/341 [============================>.] - ETA: 0s - loss: 0.5393 - accuracy: 0.7244\n",
      "Epoch 15: saving model to output/Checkpoints/undersampleModel/weights.15.hdf5\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5393 - accuracy: 0.7239\n",
      "Epoch 16/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5383 - accuracy: 0.7234\n",
      "Epoch 17/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5376 - accuracy: 0.7257\n",
      "Epoch 18/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5375 - accuracy: 0.7259\n",
      "Epoch 19/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5369 - accuracy: 0.7278\n",
      "Epoch 20/25\n",
      "338/341 [============================>.] - ETA: 0s - loss: 0.5360 - accuracy: 0.7270\n",
      "Epoch 20: saving model to output/Checkpoints/undersampleModel/weights.20.hdf5\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5359 - accuracy: 0.7271\n",
      "Epoch 21/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5357 - accuracy: 0.7270\n",
      "Epoch 22/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5358 - accuracy: 0.7264\n",
      "Epoch 23/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5353 - accuracy: 0.7272\n",
      "Epoch 24/25\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5349 - accuracy: 0.7279\n",
      "Epoch 25/25\n",
      "340/341 [============================>.] - ETA: 0s - loss: 0.5352 - accuracy: 0.7267\n",
      "Epoch 25: saving model to output/Checkpoints/undersampleModel/weights.25.hdf5\n",
      "341/341 [==============================] - 3s 8ms/step - loss: 0.5350 - accuracy: 0.7267\n"
     ]
    }
   ],
   "source": [
    "# Train the model\n",
    "fit_under_nn =\\\n",
    "    under_nn.fit(\n",
    "        X_train_scaled,\n",
    "        y_train,\n",
    "        epochs=25,\n",
    "        callbacks=[cp_callback_under]\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "114/114 - 1s - loss: 0.5596 - accuracy: 0.7108 - 620ms/epoch - 5ms/step\n",
      "114/114 [==============================] - 0s 3ms/step\n",
      "Initial Model\n",
      "Loss: 0.5596217513084412, Accuracy: 0.7108234763145447\n",
      "Precision: 0.6721574929690639, Recall: 0.8772941793392763, F1: 0.7611464968152867\n"
     ]
    }
   ],
   "source": [
    "# Evaluate the model using the test data\n",
    "model_loss, model_accuracy =\\\n",
    "    under_nn.evaluate(\n",
    "        X_test_scaled,\n",
    "        y_test,\n",
    "        verbose=2,\n",
    "    )\n",
    "\n",
    "# get precision recall and f1 scores\n",
    "y_pred = under_nn.predict(X_test_scaled)\n",
    "y_pred = np.where(y_pred >= 0.5, 1, 0)\n",
    "\n",
    "precision = precision_score(y_test, y_pred)\n",
    "recall = recall_score(y_test, y_pred)\n",
    "f1 = f1_score(y_test, y_pred)\n",
    "\n",
    "\n",
    "print(\"Initial Model\")\n",
    "print(f\"Loss: {model_loss}, Accuracy: {model_accuracy}\")\n",
    "print(f\"Precision: {precision}, Recall: {recall}, F1: {f1}\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Model 4 Summary\n",
    "\n",
    "Undersmapling the T3 application_type didnt seem to have the desired effect either. It seems training the data without the T3 values was best preparing the model for predicting charity outcome success. "
   ]
  }
 ],
 "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": "3.10.9"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}