Donors_Choose_Model_2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3uxEHt2lRx5B"
   },
   "source": [
    "# Donors Choose - Model 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "YPJ_eE6rRx5D",
    "outputId": "3bbd45d7-7c98-464a-d5da-ee9ef3595cc4"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "# importing required libraries\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from keras.layers import Input, Embedding, LSTM, Dropout, BatchNormalization, Dense, concatenate, Flatten, Conv1D, MaxPool1D, LeakyReLU, ELU, SpatialDropout1D, MaxPooling1D, GlobalAveragePooling1D, GlobalMaxPooling1D\n",
    "from keras.preprocessing.text import Tokenizer, one_hot\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.models import Model, load_model\n",
    "from keras.utils import np_utils\n",
    "from keras import regularizers\n",
    "from keras.optimizers import *\n",
    "from keras.callbacks import ModelCheckpoint, EarlyStopping, TensorBoard, ReduceLROnPlateau\n",
    "\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import re\n",
    "from tqdm import tqdm\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "import seaborn as sns\n",
    "import pickle\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from scipy.sparse import hstack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 122
    },
    "colab_type": "code",
    "id": "tUlLAHBeR4nV",
    "outputId": "ba7d168d-9e21-4cda-d04b-731d04d4a939"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=email%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdocs.test%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive.photos.readonly%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fpeopleapi.readonly&response_type=code\n",
      "\n",
      "Enter your authorization code:\n",
      "··········\n",
      "Mounted at /content/drive\n"
     ]
    }
   ],
   "source": [
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 683
    },
    "colab_type": "code",
    "id": "Bjnwb2kDRx5I",
    "outputId": "6c6bad75-b40a-41e7-f054-90c5d098b05d"
   },
   "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>Unnamed: 0</th>\n",
       "      <th>Unnamed: 0.1</th>\n",
       "      <th>id</th>\n",
       "      <th>teacher_id</th>\n",
       "      <th>teacher_prefix</th>\n",
       "      <th>school_state</th>\n",
       "      <th>project_submitted_datetime</th>\n",
       "      <th>project_grade_category</th>\n",
       "      <th>project_subject_categories</th>\n",
       "      <th>project_subject_subcategories</th>\n",
       "      <th>project_title</th>\n",
       "      <th>project_essay_1</th>\n",
       "      <th>project_essay_2</th>\n",
       "      <th>project_essay_3</th>\n",
       "      <th>project_essay_4</th>\n",
       "      <th>project_resource_summary</th>\n",
       "      <th>teacher_number_of_previously_posted_projects</th>\n",
       "      <th>project_is_approved</th>\n",
       "      <th>essay</th>\n",
       "      <th>price</th>\n",
       "      <th>quantity</th>\n",
       "      <th>std_price</th>\n",
       "      <th>nrm_price</th>\n",
       "      <th>project_summary_numerical</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>160221</td>\n",
       "      <td>p253737</td>\n",
       "      <td>c90749f5d961ff158d4b4d1e7dc665fc</td>\n",
       "      <td>mrs</td>\n",
       "      <td>in</td>\n",
       "      <td>2016-12-05 13:43:57</td>\n",
       "      <td>grades_prek_2</td>\n",
       "      <td>literacy_language</td>\n",
       "      <td>esl_literacy</td>\n",
       "      <td>educational support english learners home</td>\n",
       "      <td>My students are English learners that are work...</td>\n",
       "      <td>\\\"The limits of your language are the limits o...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>My students need opportunities to practice beg...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>students english learners working english seco...</td>\n",
       "      <td>154.60</td>\n",
       "      <td>23</td>\n",
       "      <td>-0.390533</td>\n",
       "      <td>0.015397</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>140945</td>\n",
       "      <td>p258326</td>\n",
       "      <td>897464ce9ddc600bced1151f324dd63a</td>\n",
       "      <td>mr</td>\n",
       "      <td>fl</td>\n",
       "      <td>2016-10-25 09:22:10</td>\n",
       "      <td>grades_6_8</td>\n",
       "      <td>history_civics_health_sports</td>\n",
       "      <td>civics_government_teamsports</td>\n",
       "      <td>wanted projector hungry learners</td>\n",
       "      <td>Our students arrive to our school eager to lea...</td>\n",
       "      <td>The projector we need for our school is very c...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>My students need a projector to help with view...</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>students arrive school eager learn polite gene...</td>\n",
       "      <td>299.00</td>\n",
       "      <td>1</td>\n",
       "      <td>0.002396</td>\n",
       "      <td>0.029839</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>21895</td>\n",
       "      <td>p182444</td>\n",
       "      <td>3465aaf82da834c0582ebd0ef8040ca0</td>\n",
       "      <td>ms</td>\n",
       "      <td>az</td>\n",
       "      <td>2016-08-31 12:03:56</td>\n",
       "      <td>grades_6_8</td>\n",
       "      <td>health_sports</td>\n",
       "      <td>health_wellness_teamsports</td>\n",
       "      <td>soccer equipment awesome middle school students</td>\n",
       "      <td>\\r\\n\\\"True champions aren't always the ones th...</td>\n",
       "      <td>The students on the campus come to school know...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>My students need shine guards, athletic socks,...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>true champions not always ones win guts mia ha...</td>\n",
       "      <td>516.85</td>\n",
       "      <td>22</td>\n",
       "      <td>0.595191</td>\n",
       "      <td>0.051628</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>45</td>\n",
       "      <td>p246581</td>\n",
       "      <td>f3cb9bffbba169bef1a77b243e620b60</td>\n",
       "      <td>mrs</td>\n",
       "      <td>ky</td>\n",
       "      <td>2016-10-06 21:16:17</td>\n",
       "      <td>grades_prek_2</td>\n",
       "      <td>literacy_language_math_science</td>\n",
       "      <td>literacy_mathematics</td>\n",
       "      <td>techie kindergarteners</td>\n",
       "      <td>I work at a unique school filled with both ESL...</td>\n",
       "      <td>My students live in high poverty conditions wi...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>My students need to engage in Reading and Math...</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>work unique school filled esl english second l...</td>\n",
       "      <td>232.90</td>\n",
       "      <td>4</td>\n",
       "      <td>-0.177469</td>\n",
       "      <td>0.023228</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>172407</td>\n",
       "      <td>p104768</td>\n",
       "      <td>be1f7507a41f8479dc06f047086a39ec</td>\n",
       "      <td>mrs</td>\n",
       "      <td>tx</td>\n",
       "      <td>2016-07-11 01:10:09</td>\n",
       "      <td>grades_prek_2</td>\n",
       "      <td>math_science</td>\n",
       "      <td>mathematics</td>\n",
       "      <td>interactive math tools</td>\n",
       "      <td>Our second grade classroom next year will be m...</td>\n",
       "      <td>For many students, math is a subject that does...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>My students need hands on practice in mathemat...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>second grade classroom next year made around 2...</td>\n",
       "      <td>67.98</td>\n",
       "      <td>4</td>\n",
       "      <td>-0.626236</td>\n",
       "      <td>0.006733</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0  Unnamed: 0.1  ... nrm_price project_summary_numerical\n",
       "0           0        160221  ...  0.015397                         0\n",
       "1           1        140945  ...  0.029839                         0\n",
       "2           2         21895  ...  0.051628                         0\n",
       "3           3            45  ...  0.023228                         0\n",
       "4           4        172407  ...  0.006733                         0\n",
       "\n",
       "[5 rows x 24 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# reading datasets\n",
    "project_data = pd.read_csv(\"drive/My Drive/ML_data/preprocessed_data.csv\")\n",
    "project_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "id": "CwRQqVzjRx5O",
    "outputId": "ee4a28f6-363e-4c98-f61d-4d17455a93aa"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1    92706\n",
       "0    16542\n",
       "Name: project_is_approved, dtype: int64"
      ]
     },
     "execution_count": 5,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "project_data[\"project_is_approved\"].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 170
    },
    "colab_type": "code",
    "id": "O1X_5TYZRx5T",
    "outputId": "88cf66bd-31bf-4ad6-c5d2-c469eca60ad4"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Unnamed: 0', 'Unnamed: 0.1', 'id', 'teacher_id', 'teacher_prefix',\n",
       "       'school_state', 'project_submitted_datetime', 'project_grade_category',\n",
       "       'project_subject_categories', 'project_subject_subcategories',\n",
       "       'project_title', 'project_essay_1', 'project_essay_2',\n",
       "       'project_essay_3', 'project_essay_4', 'project_resource_summary',\n",
       "       'teacher_number_of_previously_posted_projects', 'project_is_approved',\n",
       "       'essay', 'price', 'quantity', 'std_price', 'nrm_price',\n",
       "       'project_summary_numerical'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 6,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "project_data.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 442
    },
    "colab_type": "code",
    "id": "RvOFOJfMRx5Y",
    "outputId": "865eff7f-c5f1-4c01-df52-45e8407b0ff7"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Unnamed: 0                                           0\n",
       "Unnamed: 0.1                                         0\n",
       "id                                                   0\n",
       "teacher_id                                           0\n",
       "teacher_prefix                                       0\n",
       "school_state                                         0\n",
       "project_submitted_datetime                           0\n",
       "project_grade_category                               0\n",
       "project_subject_categories                           0\n",
       "project_subject_subcategories                        0\n",
       "project_title                                       43\n",
       "project_essay_1                                      0\n",
       "project_essay_2                                      0\n",
       "project_essay_3                                 105490\n",
       "project_essay_4                                 105490\n",
       "project_resource_summary                             0\n",
       "teacher_number_of_previously_posted_projects         0\n",
       "project_is_approved                                  0\n",
       "essay                                                0\n",
       "price                                                0\n",
       "quantity                                             0\n",
       "std_price                                            0\n",
       "nrm_price                                            0\n",
       "project_summary_numerical                            0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# checking for null values\n",
    "project_data.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DtYhSWb1Rx5f"
   },
   "outputs": [],
   "source": [
    "# filling the null values with ''\n",
    "project_data['project_title'] = project_data['project_title'].fillna('')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "P0ZLp10cRx5k"
   },
   "outputs": [],
   "source": [
    "# combining essay and project_title columns\n",
    "project_data['cleaned_text'] = project_data['essay'] + project_data['project_title']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "k2Gv7EP7Rx5p"
   },
   "outputs": [],
   "source": [
    "# dropping unecessary columns\n",
    "project_data = project_data.drop(['Unnamed: 0', 'Unnamed: 0.1', 'id', 'teacher_id', 'project_submitted_datetime', \n",
    "                                  'project_resource_summary', 'std_price', 'nrm_price', 'project_essay_1', 'project_essay_2', \n",
    "                                  'project_essay_3', 'project_essay_4', 'essay', 'project_title'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 102
    },
    "colab_type": "code",
    "id": "HbhF48wZRx5u",
    "outputId": "3f28a28a-5da0-43f0-987a-0c977271c19a"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['teacher_prefix', 'school_state', 'project_grade_category',\n",
       "       'project_subject_categories', 'project_subject_subcategories',\n",
       "       'teacher_number_of_previously_posted_projects', 'project_is_approved',\n",
       "       'price', 'quantity', 'project_summary_numerical', 'cleaned_text'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 11,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# columns left after dropping unecessary columns\n",
    "project_data.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "sHWDzvt9Rx50"
   },
   "outputs": [],
   "source": [
    "target = project_data['project_is_approved']\n",
    "features = project_data.drop(['project_is_approved'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "eKoMQkxQRx55"
   },
   "outputs": [],
   "source": [
    "# splitting the dataset into train(75%) and test(25%) set\n",
    "X_train, X_test, y_train, y_test = train_test_split(features, target, stratify=target, test_size=0.25, random_state=123)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "2xgW9G5hRx6C",
    "outputId": "880de2c0-1364-4bcd-988c-9e5f58a048af"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of Train data (81936, 10)\n",
      "Shape of Test data (27312, 10)\n"
     ]
    }
   ],
   "source": [
    "print('Shape of Train data', X_train.shape)\n",
    "print('Shape of Test data', X_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "sNIDi36GRx6Q"
   },
   "source": [
    "**Filtering Text Data (essays & project_title) based on idf values**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "k9wWRx9QRx6Y"
   },
   "outputs": [],
   "source": [
    "tfidf = TfidfVectorizer()\n",
    "combine_tfidf = tfidf.fit_transform(X_train['cleaned_text'])\n",
    "\n",
    "# converting to dictionary\n",
    "combine_dict = dict(zip(tfidf.get_feature_names(),list(tfidf.idf_)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "LmZdLzU2Rx6w"
   },
   "outputs": [],
   "source": [
    "tfidf_df = pd.DataFrame(list(combine_dict.items()), columns=['Words', 'IDF Values'])\n",
    "tfidf_df = tfidf_df.sort_values(by ='IDF Values' )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "Ipll4kuURx6-",
    "outputId": "eb89ad57-33a1-4df8-c55e-f0909801dc1a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.0074234116042697\n",
      "11.62055875771544\n"
     ]
    }
   ],
   "source": [
    "# finding the min & max idf values\n",
    "print(tfidf_df['IDF Values'].min())\n",
    "print(tfidf_df['IDF Values'].max())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "BU2eJIlyRx7M",
    "outputId": "da0d36d1-4b62-4218-8243-82ed78d74abf"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(27769, 2)"
      ]
     },
     "execution_count": 18,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# based on the idf values we prepare the corpus, thereby leaving the words with lower idf value than 2 and greater than 11\n",
    "corpus = tfidf_df[(tfidf_df['IDF Values'] >= 2) & (tfidf_df['IDF Values'] <=11 )]\n",
    "corpus.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 187
    },
    "colab_type": "code",
    "id": "18gpfrjXRx7f",
    "outputId": "ead4fa20-4598-4f48-acb6-fd9362c2bea2"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['new',\n",
       " 'year',\n",
       " 'one',\n",
       " 'would',\n",
       " 'time',\n",
       " 'student',\n",
       " 'want',\n",
       " 'skills',\n",
       " 'grade',\n",
       " 'reading']"
      ]
     },
     "execution_count": 19,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vocab = corpus[\"Words\"].tolist()\n",
    "vocab[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "15XLMpGbRx7x"
   },
   "source": [
    "**Tokenizing the Text**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "LxQ4F2_XRx70"
   },
   "outputs": [],
   "source": [
    "# convert the sentences (strings) into integers\n",
    "tokenizer = Tokenizer()\n",
    "tokenizer.fit_on_texts(vocab)\n",
    "sequences_train = tokenizer.texts_to_sequences(X_train['cleaned_text'])\n",
    "sequences_test = tokenizer.texts_to_sequences(X_test['cleaned_text'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "y-YKt4LuRx8G",
    "outputId": "3e6dbcf1-68d8-4b55-cd62-68194702ff7f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 27769 unique tokens.\n"
     ]
    }
   ],
   "source": [
    "# get word -> integer mapping\n",
    "word2idx = tokenizer.word_index\n",
    "print('Found %s unique tokens.' % len(word2idx))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "xNXd7b9lRx8f"
   },
   "source": [
    "**Padding the sequences**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Xm5e4eWFRx8g"
   },
   "source": [
    "The sequences have different lengths and Keras prefers inputs to be vectorized and all inputs to have the same length. \n",
    "We will pad all input sequences to have the length of 250"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "457nWHSTRx8j",
    "outputId": "56804777-5a23-4875-bb56-e5b25dbb2448"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of data tensor: (81936, 250)\n"
     ]
    }
   ],
   "source": [
    "encoded_train = pad_sequences(sequences_train, maxlen=250, padding='post', truncating='post')\n",
    "print('Shape of data tensor:', encoded_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "dtD1FSfERx9B",
    "outputId": "55f7b028-9339-4bdc-814a-4ede748a501f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of data tensor: (27312, 250)\n"
     ]
    }
   ],
   "source": [
    "encoded_test = pad_sequences(sequences_test, maxlen=250,padding='post', truncating='post')\n",
    "print('Shape of data tensor:', encoded_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "FfhB6e54Rx9K"
   },
   "source": [
    "**Getting the vector representation using Glove vectors**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "w5F0VaCERx9N"
   },
   "outputs": [],
   "source": [
    "# Loading Embedding File\n",
    "pickle_in = open('drive/My Drive/ML_data/glove_vectors','rb')\n",
    "glove_words = pickle.load(pickle_in)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "y9ibo_7wRx9w"
   },
   "outputs": [],
   "source": [
    "num_words = len(word2idx) + 1\n",
    "embedding_matrix = np.zeros((num_words, 300))\n",
    "for word, i in word2idx.items():\n",
    "    if i < len(vocab):\n",
    "        embedding_vector = glove_words.get(word)\n",
    "        if embedding_vector is not None:\n",
    "          # words not found in embedding index will be all zeros.\n",
    "          embedding_matrix[i] = embedding_vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "id": "ThjkpQymRx99",
    "outputId": "20057e93-f35c-40a5-9af4-e827cec39a5e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "27770\n",
      "--------------\n",
      "(27770, 300)\n"
     ]
    }
   ],
   "source": [
    "print(num_words)\n",
    "print('--------------')\n",
    "print(embedding_matrix.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "z4_kJjjVRx-K"
   },
   "source": [
    "**Vectorizing all the categorical features using CountVectorizer**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_Ba_ZQz0Rx-M"
   },
   "outputs": [],
   "source": [
    "vect = CountVectorizer(binary=True)\n",
    "\n",
    "train_prefix = vect.fit_transform(X_train[\"teacher_prefix\"])\n",
    "test_prefix = vect.transform(X_test[\"teacher_prefix\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "IEg1gJEaRx-S"
   },
   "outputs": [],
   "source": [
    "vect = CountVectorizer(binary=True)\n",
    "\n",
    "train_state = vect.fit_transform(X_train[\"school_state\"])\n",
    "test_state = vect.transform(X_test[\"school_state\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FNo47vTmRx-Z"
   },
   "outputs": [],
   "source": [
    "vect = CountVectorizer(binary=True)\n",
    "\n",
    "train_grade = vect.fit_transform(X_train[\"project_grade_category\"])\n",
    "test_grade = vect.transform(X_test[\"project_grade_category\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_xCF7BFyRx-m"
   },
   "outputs": [],
   "source": [
    "vect = CountVectorizer(binary=True)\n",
    "\n",
    "train_subcat = vect.fit_transform(X_train[\"project_subject_categories\"])\n",
    "test_subcat = vect.transform(X_test[\"project_subject_categories\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "jUhVbBWdRx_C"
   },
   "outputs": [],
   "source": [
    "vect = CountVectorizer(binary=True)\n",
    "\n",
    "train_subcat_1 = vect.fit_transform(X_train[\"project_subject_subcategories\"])\n",
    "test_subcat_1 = vect.transform(X_test[\"project_subject_subcategories\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "mxsxKFFfRx_Z"
   },
   "source": [
    "**Reshaping & Standardizing numerical features**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_QRPax6ERx_b"
   },
   "outputs": [],
   "source": [
    "num_train_1=X_train['project_summary_numerical'].values.reshape(-1, 1)\n",
    "num_train_2=X_train['price'].values.reshape(-1, 1)\n",
    "num_train_3=X_train['quantity'].values.reshape(-1, 1)\n",
    "num_train_4=X_train['teacher_number_of_previously_posted_projects'].values.reshape(-1, 1)\n",
    "\n",
    "num_test_1=X_test['project_summary_numerical'].values.reshape(-1, 1)\n",
    "num_test_2=X_test['price'].values.reshape(-1, 1)\n",
    "num_test_3=X_test['quantity'].values.reshape(-1, 1)\n",
    "num_test_4=X_test['teacher_number_of_previously_posted_projects'].values.reshape(-1, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "a2kTxZASRx_o"
   },
   "outputs": [],
   "source": [
    "# concatenating train numerical features\n",
    "num_train=np.concatenate((num_train_1,num_train_2,num_train_3,num_train_4),axis=1)\n",
    "\n",
    "# concatenating test numerical features\n",
    "num_test=np.concatenate((num_test_1,num_test_2,num_test_3,num_test_4),axis=1)\n",
    "\n",
    "# Standardizing the features\n",
    "norm=StandardScaler()\n",
    "norm_train=norm.fit_transform(num_train)\n",
    "norm_test=norm.transform(num_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "pdiTYTIjRx_w"
   },
   "outputs": [],
   "source": [
    "# concatencating categorical features\n",
    "cat_train = hstack([train_prefix,train_state,train_grade,train_subcat,train_subcat_1]).todense()\n",
    "cat_test = hstack([test_prefix,test_state,test_grade,test_subcat,test_subcat_1]).todense()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "rA0WoWBbRx_9"
   },
   "outputs": [],
   "source": [
    "# concatenating the numerical & categorical features\n",
    "all_train = np.hstack((cat_train,norm_train))\n",
    "all_test = np.hstack((cat_test,norm_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "sjI9fjd6RyAS"
   },
   "outputs": [],
   "source": [
    "final_train = np.expand_dims(all_train,2)\n",
    "final_test = np.expand_dims(all_test,2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "id": "fF2SHbf7RyAc",
    "outputId": "189ab8d8-604c-48f7-96f4-785ebd87db9b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(81936, 512, 1)\n",
      "----------------\n",
      "(27312, 512, 1)\n"
     ]
    }
   ],
   "source": [
    "print(final_train.shape)\n",
    "print('----------------')\n",
    "print(final_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "GapQXIB2RyAj"
   },
   "source": [
    "## Defining model architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 258
    },
    "colab_type": "code",
    "id": "W8jlxZhvRyAs",
    "outputId": "f37587ad-4f0c-4b99-cb07-0e4ac4df5520"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:66: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
      "\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:541: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
      "\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4432: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
      "\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:190: The name tf.get_default_session is deprecated. Please use tf.compat.v1.get_default_session instead.\n",
      "\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:197: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.\n",
      "\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3733: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
     ]
    }
   ],
   "source": [
    "# load pre-trained word embeddings into an Embedding layer\n",
    "# note that we set trainable = False so as to keep the embeddings of fixed sized\n",
    "embedding_layer = Embedding(\n",
    "  num_words,\n",
    "  300,\n",
    "  weights=[embedding_matrix],\n",
    "  input_length=250,\n",
    "  trainable=False\n",
    ")\n",
    "input_text = Input(shape=(250,),name=\"input_text\")\n",
    "x = embedding_layer(input_text)\n",
    "x = LSTM(100,recurrent_dropout=0.5,kernel_regularizer=regularizers.l2(0.001),return_sequences=True)(x)\n",
    "flatten_1 = Flatten()(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "nJvmJIZdRyA6",
    "outputId": "0df85341-79fb-4396-cee3-06de1c03354c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unique Categories: 5 Embedding Size: 3\n"
     ]
    }
   ],
   "source": [
    "# Now will prepare all the remaining categorical features\n",
    "# Teacher Prefix\n",
    "no_of_unique_prefix  = X_train[\"teacher_prefix\"].nunique()\n",
    "embedding_size_prefix = int(min(np.ceil((no_of_unique_prefix)/2), 50 ))\n",
    "print('Unique Categories:', no_of_unique_prefix, 'Embedding Size:', embedding_size_prefix)\n",
    "\n",
    "\n",
    "# Defining Input and Embedding Layer for the same\n",
    "\n",
    "input_prefix = Input(shape=(1,), name=\"teacher_prefix\")\n",
    "embedding_prefix = Embedding(no_of_unique_prefix, embedding_size_prefix, name=\"emb_pre\", trainable=True)(input_prefix)\n",
    "flatten_2 = Flatten()(embedding_prefix)\n",
    "\n",
    "lb = LabelEncoder()\n",
    "encoder_prefix_train = lb.fit_transform(X_train[\"teacher_prefix\"])\n",
    "encoder_prefix_test = lb.transform(X_test[\"teacher_prefix\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "VgYCJzwtRyBS",
    "outputId": "aa7b72b6-f198-4743-e28c-e2e943ecaa78"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unique Categories: 51 Embedding Size: 26\n"
     ]
    }
   ],
   "source": [
    "# School State\n",
    "no_of_unique_state  = X_train[\"school_state\"].nunique()\n",
    "embedding_size_state= int(min(np.ceil((no_of_unique_state)/2), 50 ))\n",
    "print('Unique Categories:', no_of_unique_state, 'Embedding Size:', embedding_size_state)\n",
    "\n",
    "\n",
    "# Defining Input and Embedding Layer for the same\n",
    "\n",
    "input_state = Input(shape=(1,), name=\"school_prefix\")\n",
    "embedding_state = Embedding(no_of_unique_state, embedding_size_state, name=\"emb_state\", trainable=True)(input_state)\n",
    "flatten_3 = Flatten()(embedding_state)\n",
    "\n",
    "\n",
    "encoder_state_train = lb.fit_transform(X_train[\"school_state\"])\n",
    "# encoder_state_cv = lb.transform(X_cv[\"school_state\"])\n",
    "encoder_state_test = lb.transform(X_test[\"school_state\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "KgYrb8u_RyBl",
    "outputId": "4312fba8-32a7-49ac-b6ee-ef5481096f06"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unique Categories: 4 Embedding Size: 2\n"
     ]
    }
   ],
   "source": [
    "# For project_grade_category\n",
    "no_of_unique_grade  = X_train[\"project_grade_category\"].nunique()\n",
    "embedding_size_grade = int(min(np.ceil((no_of_unique_grade)/2), 50 ))\n",
    "print('Unique Categories:', no_of_unique_grade, 'Embedding Size:', embedding_size_grade)\n",
    "\n",
    "\n",
    "# Defining Input and Embedding Layer for the same\n",
    "\n",
    "input_grade= Input(shape=(1,),name=\"grade_cat\")\n",
    "embedding_grade = Embedding(no_of_unique_grade, embedding_size_grade, name=\"emb_grade\", trainable=True)(input_grade)\n",
    "flatten_4 = Flatten()(embedding_grade)\n",
    "\n",
    "\n",
    "encoder_grade_train = lb.fit_transform(X_train[\"project_grade_category\"])\n",
    "# encoder_grade_cv = lb.transform(X_cv[\"project_grade_category\"])\n",
    "encoder_grade_test = lb.transform(X_test[\"project_grade_category\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "LkWkYFyDRyBu",
    "outputId": "2c3b0c08-76db-4c79-b22d-135f664f1a8f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unique Categories: 51 Embedding Size: 26\n"
     ]
    }
   ],
   "source": [
    "# For project_subject_categories\n",
    "no_of_unique_subcat  = X_train[\"project_subject_categories\"].nunique()\n",
    "embedding_size_subcat = int(min(np.ceil((no_of_unique_subcat)/2), 50 ))\n",
    "print('Unique Categories:', no_of_unique_subcat,'Embedding Size:', embedding_size_subcat)\n",
    "\n",
    "\n",
    "# Defining Input and Embedding Layer for the same\n",
    "\n",
    "input_subcat= Input(shape=(1,),name=\"sub_cat\")\n",
    "embedding_subcat = Embedding(no_of_unique_subcat,embedding_size_subcat,name=\"emb_subcat\",trainable=True)(input_subcat)\n",
    "flatten_5 = Flatten()(embedding_subcat)\n",
    "\n",
    "le = LabelEncoder()\n",
    "le.fit(X_train[\"project_subject_categories\"])\n",
    "X_test[\"project_subject_categories\"] = X_test[\"project_subject_categories\"].map(lambda s: '<unknown>' if s not in le.classes_ else s)\n",
    "\n",
    "le.classes_ = np.append(le.classes_, '<unknown>')\n",
    "encoder_subcat_train = le.transform(X_train[\"project_subject_categories\"])\n",
    "encoder_subcat_test= le.transform(X_test[\"project_subject_categories\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "6XQDwI9ZRyB2",
    "outputId": "5d0de29d-e92d-46e5-f61f-70ac3d89a0db"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unique Categories: 397 Embedding Size: 50\n"
     ]
    }
   ],
   "source": [
    "# For project_subject_subcategories\n",
    "no_of_unique_subcat_1  = X_train[\"project_subject_subcategories\"].nunique()\n",
    "embedding_size_subcat_1 = int(min(np.ceil((no_of_unique_subcat_1)/2), 50 ))\n",
    "print('Unique Categories:', no_of_unique_subcat_1,'Embedding Size:', embedding_size_subcat_1)\n",
    "\n",
    "# Defining Input and Embedding Layer for the same\n",
    "\n",
    "input_subcat_1= Input(shape=(1,),name=\"sub_cat_1\")\n",
    "embedding_subcat_1 = Embedding(no_of_unique_subcat_1,embedding_size_subcat_1,name=\"emb_subcat_1\",trainable=True)(input_subcat_1)\n",
    "flatten_6 = Flatten()(embedding_subcat_1)\n",
    "\n",
    "\n",
    "le = LabelEncoder()\n",
    "le.fit(X_train[\"project_subject_subcategories\"])\n",
    "X_test[\"project_subject_subcategories\"] = X_test[\"project_subject_subcategories\"].map(lambda s: '<unknown>' if s not in le.classes_ else s)\n",
    "\n",
    "le.classes_ = np.append(le.classes_, '<unknown>')\n",
    "encoder_subcat_1_train = le.transform(X_train[\"project_subject_subcategories\"])\n",
    "encoder_subcat_1_test= le.transform(X_test[\"project_subject_subcategories\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "QGIB0t-FRyCL"
   },
   "outputs": [],
   "source": [
    "# Defining the Input and Embedding Layer for the same\n",
    "\n",
    "num_feats = Input(shape=(4,),name=\"numerical_features\")\n",
    "num_feats_ = Dense(100,activation=\"relu\",kernel_initializer=\"he_normal\")(num_feats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "Po-PKkpIRyCS",
    "outputId": "630b01e9-e5ca-4992-df93-734674d31390"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Building Model-2\n"
     ]
    }
   ],
   "source": [
    "print(\"Building Model-2\")\n",
    "x_concatenate = concatenate([flatten_1, flatten_2, flatten_3, flatten_4, flatten_5, flatten_6, num_feats_])\n",
    "\n",
    "x = Dense(128,activation=\"relu\", kernel_initializer=\"he_normal\",kernel_regularizer=regularizers.l2(0.001))(x_concatenate)\n",
    "\n",
    "x=Dropout(0.5)(x)\n",
    "\n",
    "x = Dense(256,activation=\"relu\",kernel_initializer=\"he_normal\",kernel_regularizer=regularizers.l2(0.001))(x)\n",
    "\n",
    "x=Dropout(0.5)(x)\n",
    "\n",
    "x = Dense(64,activation=\"relu\", kernel_initializer=\"he_normal\",kernel_regularizer=regularizers.l2(0.001))(x)\n",
    "\n",
    "output = Dense(2, activation='softmax', name='output')(x)\n",
    "model_2 = Model(inputs=[input_text, input_prefix, input_state, input_grade,\n",
    "                        input_subcat, input_subcat_1, num_feats],outputs=[output])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Yf0A39WbRyCa"
   },
   "outputs": [],
   "source": [
    "train_data_2 = [encoded_train,encoder_prefix_train,encoder_state_train,\n",
    "              encoder_grade_train,encoder_subcat_train,encoder_subcat_1_train,norm_train]\n",
    "test_data_2 = [encoded_test,encoder_prefix_test,encoder_state_test,encoder_grade_test,\n",
    "             encoder_subcat_test,encoder_subcat_1_test,norm_test]\n",
    "\n",
    "Y_train = np_utils.to_categorical(y_train, 2) \n",
    "Y_test = np_utils.to_categorical(y_test, 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "KhPZbOWkRyCj"
   },
   "outputs": [],
   "source": [
    "checkpoint_2 = ModelCheckpoint(\"model_2.h5\",\n",
    "                             monitor=\"val_loss\",\n",
    "                             mode=\"min\",\n",
    "                             save_best_only = True,\n",
    "                             verbose=1)\n",
    "\n",
    "earlystop_2 = EarlyStopping(monitor = 'val_loss', \n",
    "                            mode=\"min\",\n",
    "                            min_delta = 0, \n",
    "                            patience = 5,\n",
    "                            verbose = 1,\n",
    "                            restore_best_weights = True)\n",
    "\n",
    "reduce_lr_2 = ReduceLROnPlateau(monitor = 'val_loss', factor = 0.2, patience = 2, verbose = 1, min_delta = 0.0001)\n",
    "\n",
    "tensorboard_2 = TensorBoard(log_dir='graph_2', histogram_freq=0, batch_size=512, write_graph=True, write_grads=False, write_images=False, embeddings_freq=0, embeddings_layer_names=None, embeddings_metadata=None, embeddings_data=None, update_freq='epoch')\n",
    "\n",
    "callbacks_2 = [checkpoint_2,earlystop_2,tensorboard_2,reduce_lr_2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "LLQM-wOuRyCt"
   },
   "outputs": [],
   "source": [
    "# Defining Custom ROC-AUC function\n",
    "from sklearn.metrics import roc_auc_score\n",
    "\n",
    "def auc1(y_true, y_pred):\n",
    "    if len(np.unique(y_true[:,1])) == 1:\n",
    "        return 0.5\n",
    "    else:\n",
    "        return roc_auc_score(y_true, y_pred)\n",
    "\n",
    "def auroc(y_true, y_pred):\n",
    "    return tf.py_func(auc1, (y_true, y_pred), tf.double)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "whShTTtvRyDB"
   },
   "outputs": [],
   "source": [
    "adam = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 292
    },
    "colab_type": "code",
    "id": "DKeB5deuRyDQ",
    "outputId": "fe74c068-edcd-4741-b4aa-7a8479b5cb5c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:793: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
      "\n",
      "WARNING:tensorflow:From <ipython-input-48-a7e6cba44e56>:10: py_func (from tensorflow.python.ops.script_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "tf.py_func is deprecated in TF V2. Instead, there are two\n",
      "    options available in V2.\n",
      "    - tf.py_function takes a python function which manipulates tf eager\n",
      "    tensors instead of numpy arrays. It's easy to convert a tf eager tensor to\n",
      "    an ndarray (just call tensor.numpy()) but having access to eager tensors\n",
      "    means `tf.py_function`s can use accelerators such as GPUs as well as\n",
      "    being differentiable using a gradient tape.\n",
      "    - tf.numpy_function maintains the semantics of the deprecated tf.py_func\n",
      "    (it is not differentiable, and manipulates numpy arrays). It drops the\n",
      "    stateful argument making all functions stateful.\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "model_2.compile(optimizer=adam, loss='categorical_crossentropy', metrics=[auroc])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "iJoWnjwLRyDX",
    "outputId": "dc838385-2be6-4363-b7b0-a08e05995fcb",
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.where in 2.0, which has the same broadcast rule as np.where\n",
      "Train on 81936 samples, validate on 27312 samples\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/callbacks.py:1122: The name tf.summary.merge_all is deprecated. Please use tf.compat.v1.summary.merge_all instead.\n",
      "\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/callbacks.py:1125: The name tf.summary.FileWriter is deprecated. Please use tf.compat.v1.summary.FileWriter instead.\n",
      "\n",
      "Epoch 1/30\n",
      "81936/81936 [==============================] - 75s 910us/step - loss: 1.0141 - auroc: 0.6706 - val_loss: 0.6913 - val_auroc: 0.7289\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.69125, saving model to model_2.h5\n",
      "Epoch 2/30\n",
      "81936/81936 [==============================] - 73s 896us/step - loss: 0.5739 - auroc: 0.7312 - val_loss: 0.5311 - val_auroc: 0.7439\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.69125 to 0.53105, saving model to model_2.h5\n",
      "Epoch 3/30\n",
      "81936/81936 [==============================] - 73s 896us/step - loss: 0.4681 - auroc: 0.7479 - val_loss: 0.4472 - val_auroc: 0.7527\n",
      "\n",
      "Epoch 00003: val_loss improved from 0.53105 to 0.44721, saving model to model_2.h5\n",
      "Epoch 4/30\n",
      "81936/81936 [==============================] - 74s 902us/step - loss: 0.4277 - auroc: 0.7562 - val_loss: 0.4180 - val_auroc: 0.7554\n",
      "\n",
      "Epoch 00004: val_loss improved from 0.44721 to 0.41802, saving model to model_2.h5\n",
      "Epoch 5/30\n",
      "81936/81936 [==============================] - 73s 892us/step - loss: 0.4077 - auroc: 0.7598 - val_loss: 0.4179 - val_auroc: 0.7561\n",
      "\n",
      "Epoch 00005: val_loss improved from 0.41802 to 0.41786, saving model to model_2.h5\n",
      "Epoch 6/30\n",
      "81936/81936 [==============================] - 73s 894us/step - loss: 0.3976 - auroc: 0.7608 - val_loss: 0.4140 - val_auroc: 0.7596\n",
      "\n",
      "Epoch 00006: val_loss improved from 0.41786 to 0.41397, saving model to model_2.h5\n",
      "Epoch 7/30\n",
      "81936/81936 [==============================] - 74s 898us/step - loss: 0.3934 - auroc: 0.7625 - val_loss: 0.3929 - val_auroc: 0.7578\n",
      "\n",
      "Epoch 00007: val_loss improved from 0.41397 to 0.39290, saving model to model_2.h5\n",
      "Epoch 8/30\n",
      "81936/81936 [==============================] - 74s 905us/step - loss: 0.3869 - auroc: 0.7652 - val_loss: 0.3960 - val_auroc: 0.7572\n",
      "\n",
      "Epoch 00008: val_loss did not improve from 0.39290\n",
      "Epoch 9/30\n",
      "81936/81936 [==============================] - 74s 905us/step - loss: 0.3866 - auroc: 0.7661 - val_loss: 0.3919 - val_auroc: 0.7597\n",
      "\n",
      "Epoch 00009: val_loss improved from 0.39290 to 0.39190, saving model to model_2.h5\n",
      "Epoch 10/30\n",
      "81936/81936 [==============================] - 74s 906us/step - loss: 0.3849 - auroc: 0.7663 - val_loss: 0.3831 - val_auroc: 0.7598\n",
      "\n",
      "Epoch 00010: val_loss improved from 0.39190 to 0.38313, saving model to model_2.h5\n",
      "Epoch 11/30\n",
      "81936/81936 [==============================] - 75s 911us/step - loss: 0.3820 - auroc: 0.7676 - val_loss: 0.3885 - val_auroc: 0.7600\n",
      "\n",
      "Epoch 00011: val_loss did not improve from 0.38313\n",
      "Epoch 12/30\n",
      "81936/81936 [==============================] - 75s 910us/step - loss: 0.3813 - auroc: 0.7680 - val_loss: 0.3870 - val_auroc: 0.7607\n",
      "\n",
      "Epoch 00012: val_loss did not improve from 0.38313\n",
      "\n",
      "Epoch 00012: ReduceLROnPlateau reducing learning rate to 0.00020000000949949026.\n",
      "Epoch 13/30\n",
      "81936/81936 [==============================] - 74s 904us/step - loss: 0.3723 - auroc: 0.7755 - val_loss: 0.3796 - val_auroc: 0.7629\n",
      "\n",
      "Epoch 00013: val_loss improved from 0.38313 to 0.37965, saving model to model_2.h5\n",
      "Epoch 14/30\n",
      "81936/81936 [==============================] - 74s 902us/step - loss: 0.3673 - auroc: 0.7806 - val_loss: 0.3788 - val_auroc: 0.7627\n",
      "\n",
      "Epoch 00014: val_loss improved from 0.37965 to 0.37884, saving model to model_2.h5\n",
      "Epoch 15/30\n",
      "81936/81936 [==============================] - 74s 906us/step - loss: 0.3660 - auroc: 0.7828 - val_loss: 0.3781 - val_auroc: 0.7641\n",
      "\n",
      "Epoch 00015: val_loss improved from 0.37884 to 0.37815, saving model to model_2.h5\n",
      "Epoch 16/30\n",
      "81936/81936 [==============================] - 75s 914us/step - loss: 0.3644 - auroc: 0.7855 - val_loss: 0.3800 - val_auroc: 0.7616\n",
      "\n",
      "Epoch 00016: val_loss did not improve from 0.37815\n",
      "Epoch 17/30\n",
      "81936/81936 [==============================] - 74s 908us/step - loss: 0.3635 - auroc: 0.7875 - val_loss: 0.3804 - val_auroc: 0.7624\n",
      "\n",
      "Epoch 00017: val_loss did not improve from 0.37815\n",
      "\n",
      "Epoch 00017: ReduceLROnPlateau reducing learning rate to 4.0000001899898055e-05.\n",
      "Epoch 18/30\n",
      "81936/81936 [==============================] - 74s 903us/step - loss: 0.3594 - auroc: 0.7928 - val_loss: 0.3804 - val_auroc: 0.7625\n",
      "\n",
      "Epoch 00018: val_loss did not improve from 0.37815\n",
      "Epoch 19/30\n",
      "81936/81936 [==============================] - 74s 902us/step - loss: 0.3575 - auroc: 0.7952 - val_loss: 0.3807 - val_auroc: 0.7623\n",
      "\n",
      "Epoch 00019: val_loss did not improve from 0.37815\n",
      "\n",
      "Epoch 00019: ReduceLROnPlateau reducing learning rate to 8.000000525498762e-06.\n",
      "Epoch 20/30\n",
      "81936/81936 [==============================] - 74s 903us/step - loss: 0.3565 - auroc: 0.7959 - val_loss: 0.3809 - val_auroc: 0.7622\n",
      "\n",
      "Epoch 00020: val_loss did not improve from 0.37815\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00020: early stopping\n"
     ]
    }
   ],
   "source": [
    "history_2 = model_2.fit(train_data_2, Y_train, batch_size=512,\n",
    "                        epochs=30, validation_data=(test_data_2,Y_test), callbacks=callbacks_2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 531
    },
    "colab_type": "code",
    "id": "hgkj62AMRyDg",
    "outputId": "033c1d0d-6ffd-409a-ffe3-ff322b8452bc"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEBCAYAAACXArmGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEMLt/wPgTnC7noeCSgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAEBCAYAAAB/rs7oAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIWKMJH4hhIgxkviFECLG/H9YZSl/eKrh5QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set()\n",
    "plt.plot(history_2.history['auroc'], 'r')\n",
    "plt.plot(history_2.history['val_auroc'], 'b')\n",
    "plt.legend({'Train ROCAUC': 'r', 'Test ROCAUC':'b'})\n",
    "plt.show()\n",
    "\n",
    "\n",
    "plt.plot(history_2.history['loss'], 'r')\n",
    "plt.plot(history_2.history['val_loss'], 'b')\n",
    "plt.legend({'Train Loss': 'r', 'Test Loss':'b'})\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "zySEalj_RyDt"
   },
   "source": [
    "### AUC Score for Model 2  - 0.7622"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Donors Choose - Model 2.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}