Full_Naive_Bayes_Analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "yzhXj29rU8Bi"
   },
   "source": [
    "# Naive Bayes NB - Algorithm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "25ZXYCJoU8Bj"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "raw",
    "id": "pUBhd-4lU8Bm"
   },
   "source": [
    "<p>\n",
    "DonorsChoose.org receives hundreds of thousands of project proposals each year for classroom projects in need of funding. Right now, a large number of volunteers is needed to manually screen each submission before it's approved to be posted on the DonorsChoose.org website.\n",
    "</p>\n",
    "<p>\n",
    "    Next year, DonorsChoose.org expects to receive close to 500,000 project proposals. As a result, there are three main problems they need to solve:\n",
    "<ul>\n",
    "<li>\n",
    "    How to scale current manual processes and resources to screen 500,000 projects so that they can be posted as quickly and as efficiently as possible</li>F\n",
    "    <li>How to increase the consistency of project vetting across different volunteers to improve the experience for teachers</li>\n",
    "    <li>How to focus volunteer time on the applications that need the most assistance</li>\n",
    "    </ul>\n",
    "</p>    \n",
    "<p>\n",
    "The goal of the competition is to predict whether or not a DonorsChoose.org project proposal submitted by a teacher will be approved, using the text of project descriptions as well as additional metadata about the project, teacher, and school. DonorsChoose.org can then use this information to identify projects most likely to need further review before approval.\n",
    "</p>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "HJ0MYvNUU8Bn"
   },
   "source": [
    "### Notes on the Essay Data\n",
    "\n",
    "<ul>\n",
    "Prior to May 17, 2016, the prompts for the essays were as follows:\n",
    "<li>__project_essay_1:__ \"Introduce us to your classroom\"</li>\n",
    "<li>__project_essay_2:__ \"Tell us more about your students\"</li>\n",
    "<li>__project_essay_3:__ \"Describe how your students will use the materials you're requesting\"</li>\n",
    "<li>__project_essay_3:__ \"Close by sharing why your project will make a difference\"</li>\n",
    "</ul>\n",
    "\n",
    "\n",
    "<ul>\n",
    "Starting on May 17, 2016, the number of essays was reduced from 4 to 2, and the prompts for the first 2 essays were changed to the following:<br>\n",
    "<li>__project_essay_1:__ \"Describe your students: What makes your students special? Specific details about their background, your neighborhood, and your school are all helpful.\"</li>\n",
    "<li>__project_essay_2:__ \"About your project: How will these materials make a difference in your students' learning and improve their school lives?\"</li>\n",
    "<br>For all projects with project_submitted_datetime of 2016-05-17 and later, the values of project_essay_3 and project_essay_4 will be NaN.\n",
    "</ul>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FmDp68RVVTCY"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "NzelmVocVlQf"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "t7NxPVu2VlnC"
   },
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "PwLN0KxBU8Bo"
   },
   "source": [
    "## Step 1: Import the necessary Libraries\n",
    "we will need to import libraries that allow for data analysis and data visualization to get acclimated to the dataset. We will be using pandas, numpy, matplotlib and seaborn to conduct this. Data Exploration libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "SpUqG1QpU8Bp"
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "warnings.filterwarnings(action='ignore', category=UserWarning, module='gensim')\n",
    "warnings.filterwarnings(\"ignore\",'detected Windows; aliasing chunkize to chunkize_serial')\n",
    "warnings.filterwarnings(\"ignore\", message=\"numpy.dtype size changed\")\n",
    "\n",
    "import sqlite3\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import nltk\n",
    "import string\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "from sklearn.feature_extraction.text import TfidfTransformer\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn import metrics\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "from nltk.stem.porter import PorterStemmer\n",
    "\n",
    "import re\n",
    "# Tutorial about Python regular expressions: https://pymotw.com/2/re/\n",
    "import string\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.stem import PorterStemmer\n",
    "from nltk.stem.wordnet import WordNetLemmatizer\n",
    "\n",
    "from gensim.models import Word2Vec\n",
    "from gensim.models import KeyedVectors\n",
    "import pickle\n",
    "from tqdm import tqdm\n",
    "import os\n",
    "from collections import Counter"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "HamSbnwmU8Br"
   },
   "source": [
    "## Step 2: Read in the dataset.\n",
    "We will use the pandas .read_csv() method to read in the dataset. Then we will use the. head() method to observe the first few rows of the data, to understand the information better. In our case, the feature(column) headers tell us pretty little. This is fine because we are merely trying to gain insight via classifying new data points by referencing it’s neighboring elements."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3HE2JgUhU8Bs"
   },
   "source": [
    "## 1.1 Reading Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "b8yH4nMNW-jG"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "sWEbokhZU8Bt"
   },
   "outputs": [],
   "source": [
    "project_data = pd.read_csv(\"C:\\\\VipinML\\\\Assignment 2\\\\Assignments_DonorsChoose_2018\\\\train_data.csv\")\n",
    "resource_data = pd.read_csv(\"C:\\\\VipinML\\Assignment 2\\\\Assignments_DonorsChoose_2018\\\\resources.csv\")\n",
    "#Limit the data for testing purpose since processing takes few hours for full set..\n",
    "\n",
    "project_data = project_data.head(75000)\n",
    "resource_data = resource_data.head(75000)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "FfgIobXv0PDL",
    "outputId": "3d0b98c4-49bc-41ca-ff8a-d2f3279bda39"
   },
   "outputs": [],
   "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": 158
    },
    "colab_type": "code",
    "id": "haLBXwVmU8Bw",
    "outputId": "ee422958-13ab-4504-cf1e-e3a5ba19177b",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of data points in train data (75000, 17)\n",
      "--------------------------------------------------\n",
      "The attributes of data : ['Unnamed: 0' 'id' 'teacher_id' 'teacher_prefix' 'school_state'\n",
      " 'project_submitted_datetime' 'project_grade_category'\n",
      " 'project_subject_categories' 'project_subject_subcategories'\n",
      " 'project_title' 'project_essay_1' 'project_essay_2' 'project_essay_3'\n",
      " 'project_essay_4' 'project_resource_summary'\n",
      " 'teacher_number_of_previously_posted_projects' 'project_is_approved']\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of data points in train data\", project_data.shape)\n",
    "print('-'*50)\n",
    "print(\"The attributes of data :\", project_data.columns.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 168
    },
    "colab_type": "code",
    "id": "HNsoChYcU8B0",
    "outputId": "ccfe3d9d-f9a7-40cc-c307-98bd316734b1"
   },
   "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>id</th>\n",
       "      <th>teacher_id</th>\n",
       "      <th>teacher_prefix</th>\n",
       "      <th>school_state</th>\n",
       "      <th>Date</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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>55660</td>\n",
       "      <td>8393</td>\n",
       "      <td>p205479</td>\n",
       "      <td>2bf07ba08945e5d8b2a3f269b2b3cfe5</td>\n",
       "      <td>Mrs.</td>\n",
       "      <td>CA</td>\n",
       "      <td>2016-04-27 00:27:36</td>\n",
       "      <td>Grades PreK-2</td>\n",
       "      <td>Math &amp; Science</td>\n",
       "      <td>Applied Sciences, Health &amp; Life Science</td>\n",
       "      <td>Engineering STEAM into the Primary Classroom</td>\n",
       "      <td>I have been fortunate enough to use the Fairy ...</td>\n",
       "      <td>My students come from a variety of backgrounds...</td>\n",
       "      <td>Each month I try to do several science or STEM...</td>\n",
       "      <td>It is challenging to develop high quality scie...</td>\n",
       "      <td>My students need STEM kits to learn critical s...</td>\n",
       "      <td>53</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Unnamed: 0       id                        teacher_id teacher_prefix  \\\n",
       "55660        8393  p205479  2bf07ba08945e5d8b2a3f269b2b3cfe5           Mrs.   \n",
       "\n",
       "      school_state                Date project_grade_category  \\\n",
       "55660           CA 2016-04-27 00:27:36          Grades PreK-2   \n",
       "\n",
       "      project_subject_categories            project_subject_subcategories  \\\n",
       "55660             Math & Science  Applied Sciences, Health & Life Science   \n",
       "\n",
       "                                      project_title  \\\n",
       "55660  Engineering STEAM into the Primary Classroom   \n",
       "\n",
       "                                         project_essay_1  \\\n",
       "55660  I have been fortunate enough to use the Fairy ...   \n",
       "\n",
       "                                         project_essay_2  \\\n",
       "55660  My students come from a variety of backgrounds...   \n",
       "\n",
       "                                         project_essay_3  \\\n",
       "55660  Each month I try to do several science or STEM...   \n",
       "\n",
       "                                         project_essay_4  \\\n",
       "55660  It is challenging to develop high quality scie...   \n",
       "\n",
       "                                project_resource_summary  \\\n",
       "55660  My students need STEM kits to learn critical s...   \n",
       "\n",
       "       teacher_number_of_previously_posted_projects  project_is_approved  \n",
       "55660                                            53                    1  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# how to replace elements in list python: https://stackoverflow.com/a/2582163/4084039\n",
    "cols = ['Date' if x=='project_submitted_datetime' else x for x in list(project_data.columns)]\n",
    "#sort dataframe based on time pandas python: https://stackoverflow.com/a/49702492/4084039\n",
    "project_data['Date'] = pd.to_datetime(project_data['project_submitted_datetime'])\n",
    "project_data.drop('project_submitted_datetime', axis=1, inplace=True)\n",
    "project_data.sort_values(by=['Date'], inplace=True)\n",
    "\n",
    "# how to reorder columns pandas python: https://stackoverflow.com/a/13148611/4084039\n",
    "project_data = project_data[cols]\n",
    "project_data.head(1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "OOrnQRPnU8B3"
   },
   "source": [
    "### Collect All Features into global List"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "3ju_RPOuU8B4"
   },
   "outputs": [],
   "source": [
    "# Create list of features to appned all the features for BoW, TFIDF and std catagories.\n",
    "features_bow =[]\n",
    "features_tfidf= []\n",
    "features_std =[]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "f7T2ORjbU8B6"
   },
   "source": [
    "## 1.2 preprocessing of `project_subject_categories`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "S9n5JhcnU8B7"
   },
   "outputs": [],
   "source": [
    "# https://stackoverflow.com/a/47091490/4084039\n",
    "import re\n",
    "def decontracted(phrase):\n",
    "    # specific\n",
    "    phrase = re.sub(r\"won't\", \"will not\", phrase)\n",
    "    phrase = re.sub(r\"can\\'t\", \"can not\", phrase)\n",
    "\n",
    "    # general\n",
    "    phrase = re.sub(r\"n\\'t\", \" not\", phrase)\n",
    "    phrase = re.sub(r\"\\'re\", \" are\", phrase)\n",
    "    phrase = re.sub(r\"\\'s\", \" is\", phrase)\n",
    "    phrase = re.sub(r\"\\'d\", \" would\", phrase)\n",
    "    phrase = re.sub(r\"\\'ll\", \" will\", phrase)\n",
    "    phrase = re.sub(r\"\\'t\", \" not\", phrase)\n",
    "    phrase = re.sub(r\"\\'ve\", \" have\", phrase)\n",
    "    phrase = re.sub(r\"\\'m\", \" am\", phrase)\n",
    "    return phrase"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 72
    },
    "colab_type": "code",
    "id": "O0mUFB66U8CR",
    "outputId": "f0df24a6-5911-45bf-d5d1-15c992dd985e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y students require multiple avenues of lesson adaptations due to learning and physical challenges.  I am always looking for songs and videos to introduce topics and then reinforce with hands-on activ\n",
      "====================================================================================================\n"
     ]
    }
   ],
   "source": [
    "sent = decontracted(project_data['project_essay_1'].values[500])\n",
    "print(sent[1:200])\n",
    "print(\"=\"*100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DS3PViUWU8CV"
   },
   "outputs": [],
   "source": [
    "def preprocesssing_cleanup(text):\n",
    "    # Combining all the above stundents \n",
    "   \n",
    "    from tqdm import tqdm\n",
    "    X_text = []\n",
    "    # tqdm is for printing the status bar\n",
    "    for sentance in tqdm(text):\n",
    "        print (sentance)\n",
    "        sent = decontracted(sentance)\n",
    "        sent = sent.replace('\\\\r', ' ')\n",
    "        sent = sent.replace('\\\\\"', ' ')\n",
    "        sent = sent.replace('\\\\n', ' ')\n",
    "        sent = re.sub('[^A-Za-z0-9]+', ' ', sent)\n",
    "        # https://gist.github.com/sebleier/554280\n",
    "        sent = ' '.join(e for e in sent.split() if e not in stopwords)\n",
    "        X_text.append(sent.lower().strip())\n",
    "       # print (X_train_preprocessed_essays)\n",
    "    return X_text\n",
    "\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "AzpM18CMU8CX"
   },
   "outputs": [],
   "source": [
    "catogories = list(project_data['project_subject_categories'].values)\n",
    "# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039\n",
    "\n",
    "# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/\n",
    "# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string\n",
    "# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python\n",
    "cat_list = []\n",
    "for i in catogories:\n",
    "    temp = \"\"\n",
    "    # consider we have text like this \"Math & Science, Warmth, Care & Hunger\"\n",
    "    for j in i.split(','): # it will split it in three parts [\"Math & Science\", \"Warmth\", \"Care & Hunger\"]\n",
    "        if 'The' in j.split(): # this will split each of the catogory based on space \"Math & Science\"=> \"Math\",\"&\", \"Science\"\n",
    "            j=j.replace('The','') # if we have the words \"The\" we are going to replace it with ''(i.e removing 'The')\n",
    "        j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:\"Math & Science\"=>\"Math&Science\"\n",
    "        temp+=j.strip()+\" \" #\" abc \".strip() will return \"abc\", remove the trailing spaces\n",
    "        temp = temp.replace('&','_') # we are replacing the & value into \n",
    "    cat_list.append(temp.strip())\n",
    "    \n",
    "project_data['clean_categories'] = cat_list\n",
    "project_data.drop(['project_subject_categories'], axis=1, inplace=True)\n",
    "\n",
    "from collections import Counter\n",
    "my_counter = Counter()\n",
    "for word in project_data['clean_categories'].values:\n",
    "    my_counter.update(word.split())\n",
    "\n",
    "cat_dict = dict(my_counter)\n",
    "sorted_cat_dict = dict(sorted(cat_dict.items(), key=lambda kv: kv[1]))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "BKMTtTRsU8Ca"
   },
   "source": [
    "## 1.3 preprocessing of `project_subject_subcategories`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "L_0ejP2aU8Cb"
   },
   "outputs": [],
   "source": [
    "sub_catogories = list(project_data['project_subject_subcategories'].values)\n",
    "# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039\n",
    "\n",
    "# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/\n",
    "# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string\n",
    "# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python\n",
    "\n",
    "sub_cat_list = []\n",
    "for i in sub_catogories:\n",
    "    temp = \"\"\n",
    "    # consider we have text like this \"Math & Science, Warmth, Care & Hunger\"\n",
    "    for j in i.split(','): # it will split it in three parts [\"Math & Science\", \"Warmth\", \"Care & Hunger\"]\n",
    "        if 'The' in j.split(): # this will split each of the catogory based on space \"Math & Science\"=> \"Math\",\"&\", \"Science\"\n",
    "            j=j.replace('The','') # if we have the words \"The\" we are going to replace it with ''(i.e removing 'The')\n",
    "        j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:\"Math & Science\"=>\"Math&Science\"\n",
    "        temp +=j.strip()+\" \"#\" abc \".strip() will return \"abc\", remove the trailing spaces\n",
    "        temp = temp.replace('&','_')\n",
    "    sub_cat_list.append(temp.strip())\n",
    "\n",
    "project_data['clean_subcategories'] = sub_cat_list\n",
    "project_data.drop(['project_subject_subcategories'], axis=1, inplace=True)\n",
    "\n",
    "# count of all the words in corpus python: https://stackoverflow.com/a/22898595/4084039\n",
    "my_counter = Counter()\n",
    "for word in project_data['clean_subcategories'].values:\n",
    "    my_counter.update(word.split())\n",
    "    \n",
    "sub_cat_dict = dict(my_counter)\n",
    "sorted_sub_cat_dict = dict(sorted(sub_cat_dict.items(), key=lambda kv: kv[1]))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "g9wfOjgrU8Cd"
   },
   "outputs": [],
   "source": [
    "teacher_cat = list(project_data['teacher_prefix'].values)\n",
    "# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039\n",
    "\n",
    "# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/\n",
    "# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string\n",
    "# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python\n",
    "cat_list = []\n",
    "for i in teacher_cat:\n",
    "    temp = \"\"\n",
    "    # consider we have text like this \"Math & Science, Warmth, Care & Hunger\"\n",
    "    j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:\"Math & Science\"=>\"Math&Science\"\n",
    "    temp+=j.strip()+\" \" #\" abc \".strip() will return \"abc\", remove the trailing spaces\n",
    "    temp = temp.replace('&','_') # we are replacing the & value into \n",
    "    cat_list.append(temp.strip())\n",
    "\n",
    "\n",
    "project_data.drop(['teacher_prefix'], axis=1, inplace=True)\n",
    "project_data['teacher_prefix'] = sub_cat_list\n",
    "    \n",
    "from collections import Counter\n",
    "my_counter = Counter()\n",
    "for word in project_data['teacher_prefix'].values:\n",
    "    my_counter.update(word.split())\n",
    "\n",
    "cat_dict = dict(my_counter)\n",
    "sorted_teacher_dict = dict(sorted(cat_dict.items(), key=lambda kv: kv[1]))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "T9fPeXrCU8Cg"
   },
   "source": [
    "## 1.3 Text preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "w2gHp5-bU8Ch"
   },
   "outputs": [],
   "source": [
    "# merge two column text dataframe: \n",
    "project_data[\"essay\"] = project_data[\"project_essay_1\"].map(str) +\\\n",
    "                        project_data[\"project_essay_2\"].map(str) + \\\n",
    "                        project_data[\"project_essay_3\"].map(str) + \\\n",
    "                        project_data[\"project_essay_4\"].map(str)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 220
    },
    "colab_type": "code",
    "id": "MgzikQksU8Cj",
    "outputId": "a724f7ca-405c-4a4f-8b04-50503011b6ea"
   },
   "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>id</th>\n",
       "      <th>teacher_id</th>\n",
       "      <th>school_state</th>\n",
       "      <th>Date</th>\n",
       "      <th>project_grade_category</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>clean_categories</th>\n",
       "      <th>clean_subcategories</th>\n",
       "      <th>teacher_prefix</th>\n",
       "      <th>essay</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>55660</td>\n",
       "      <td>8393</td>\n",
       "      <td>p205479</td>\n",
       "      <td>2bf07ba08945e5d8b2a3f269b2b3cfe5</td>\n",
       "      <td>CA</td>\n",
       "      <td>2016-04-27 00:27:36</td>\n",
       "      <td>Grades PreK-2</td>\n",
       "      <td>Engineering STEAM into the Primary Classroom</td>\n",
       "      <td>I have been fortunate enough to use the Fairy ...</td>\n",
       "      <td>My students come from a variety of backgrounds...</td>\n",
       "      <td>Each month I try to do several science or STEM...</td>\n",
       "      <td>It is challenging to develop high quality scie...</td>\n",
       "      <td>My students need STEM kits to learn critical s...</td>\n",
       "      <td>53</td>\n",
       "      <td>1</td>\n",
       "      <td>Math_Science</td>\n",
       "      <td>AppliedSciences Health_LifeScience</td>\n",
       "      <td>AppliedSciences Health_LifeScience</td>\n",
       "      <td>I have been fortunate enough to use the Fairy ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Unnamed: 0       id                        teacher_id school_state  \\\n",
       "55660        8393  p205479  2bf07ba08945e5d8b2a3f269b2b3cfe5           CA   \n",
       "\n",
       "                     Date project_grade_category  \\\n",
       "55660 2016-04-27 00:27:36          Grades PreK-2   \n",
       "\n",
       "                                      project_title  \\\n",
       "55660  Engineering STEAM into the Primary Classroom   \n",
       "\n",
       "                                         project_essay_1  \\\n",
       "55660  I have been fortunate enough to use the Fairy ...   \n",
       "\n",
       "                                         project_essay_2  \\\n",
       "55660  My students come from a variety of backgrounds...   \n",
       "\n",
       "                                         project_essay_3  \\\n",
       "55660  Each month I try to do several science or STEM...   \n",
       "\n",
       "                                         project_essay_4  \\\n",
       "55660  It is challenging to develop high quality scie...   \n",
       "\n",
       "                                project_resource_summary  \\\n",
       "55660  My students need STEM kits to learn critical s...   \n",
       "\n",
       "       teacher_number_of_previously_posted_projects  project_is_approved  \\\n",
       "55660                                            53                    1   \n",
       "\n",
       "      clean_categories                 clean_subcategories  \\\n",
       "55660     Math_Science  AppliedSciences Health_LifeScience   \n",
       "\n",
       "                           teacher_prefix  \\\n",
       "55660  AppliedSciences Health_LifeScience   \n",
       "\n",
       "                                                   essay  \n",
       "55660  I have been fortunate enough to use the Fairy ...  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "project_data.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 72
    },
    "colab_type": "code",
    "id": "Iea8MLzyU8Cm",
    "outputId": "c4d304ad-ceba-4597-e6c7-5ea6694cf13e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y students require multiple avenues of lesson adaptations due to learning and physical challenges.  I am always looking for songs and videos to introduce topics and then reinforce with hands-on activ\n",
      "y students require multiple avenues of lesson adaptations due to learning and physical challenges.  I am always looking for songs and videos to introduce topics and then reinforce with hands-on activ\n"
     ]
    }
   ],
   "source": [
    "# \\r \\n \\t remove from string python: http://texthandler.com/info/remove-line-breaks-python/\n",
    "sent = sent.replace('\\\\r', ' ')\n",
    "sent = sent.replace('\\\\\"', ' ')\n",
    "sent = sent.replace('\\\\n', ' ')\n",
    "print(sent[1:200])\n",
    "print(sent[1:200])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "qgbpJWtOU8Cp",
    "outputId": "6f6dda5f-7f0b-4c31-81fa-65024eee9f7d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y students require multiple avenues of lesson adaptations due to learning and physical challenges I am always looking for songs and videos to introduce topics and then reinforce with hands on activit\n"
     ]
    }
   ],
   "source": [
    "#remove spacial character: https://stackoverflow.com/a/5843547/4084039\n",
    "sent = re.sub('[^A-Za-z0-9]+', ' ', sent)\n",
    "print(sent[1:200])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "JXAPZXIvU8Cs"
   },
   "outputs": [],
   "source": [
    "# https://gist.github.com/sebleier/554280\n",
    "# we are removing the words from the stop words list: 'no', 'nor', 'not'\n",
    "stopwords= ['i', 'me', 'my', 'myself', 'we', 'our', 'ours', 'ourselves', 'you', \"you're\", \"you've\",\\\n",
    "            \"you'll\", \"you'd\", 'your', 'yours', 'yourself', 'yourselves', 'he', 'him', 'his', 'himself', \\\n",
    "            'she', \"she's\", 'her', 'hers', 'herself', 'it', \"it's\", 'its', 'itself', 'they', 'them', 'their',\\\n",
    "            'theirs', 'themselves', 'what', 'which', 'who', 'whom', 'this', 'that', \"that'll\", 'these', 'those', \\\n",
    "            'am', 'is', 'are', 'was', 'were', 'be', 'been', 'being', 'have', 'has', 'had', 'having', 'do', 'does', \\\n",
    "            'did', 'doing', 'a', 'an', 'the', 'and', 'but', 'if', 'or', 'because', 'as', 'until', 'while', 'of', \\\n",
    "            'at', 'by', 'for', 'with', 'about', 'against', 'between', 'into', 'through', 'during', 'before', 'after',\\\n",
    "            'above', 'below', 'to', 'from', 'up', 'down', 'in', 'out', 'on', 'off', 'over', 'under', 'again', 'further',\\\n",
    "            'then', 'once', 'here', 'there', 'when', 'where', 'why', 'how', 'all', 'any', 'both', 'each', 'few', 'more',\\\n",
    "            'most', 'other', 'some', 'such', 'only', 'own', 'same', 'so', 'than', 'too', 'very', \\\n",
    "            's', 't', 'can', 'will', 'just', 'don', \"don't\", 'should', \"should've\", 'now', 'd', 'll', 'm', 'o', 're', \\\n",
    "            've', 'y', 'ain', 'aren', \"aren't\", 'couldn', \"couldn't\", 'didn', \"didn't\", 'doesn', \"doesn't\", 'hadn',\\\n",
    "            \"hadn't\", 'hasn', \"hasn't\", 'haven', \"haven't\", 'isn', \"isn't\", 'ma', 'mightn', \"mightn't\", 'mustn',\\\n",
    "            \"mustn't\", 'needn', \"needn't\", 'shan', \"shan't\", 'shouldn', \"shouldn't\", 'wasn', \"wasn't\", 'weren', \"weren't\", \\\n",
    "            'won', \"won't\", 'wouldn', \"wouldn't\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3fKZ7Vj2U8Cw"
   },
   "source": [
    "### 1.4.3 Merging price with project_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 273
    },
    "colab_type": "code",
    "id": "123lY0FcU8Cz",
    "outputId": "80f3436c-9590-45e6-9cb7-ffcee7f9e485"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "        id   price  quantity\n",
      "1  p000031  357.98         2\n",
      "2  p000052  114.98         2\n"
     ]
    },
    {
     "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>id</th>\n",
       "      <th>teacher_id</th>\n",
       "      <th>school_state</th>\n",
       "      <th>Date</th>\n",
       "      <th>project_grade_category</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>clean_categories</th>\n",
       "      <th>clean_subcategories</th>\n",
       "      <th>teacher_prefix</th>\n",
       "      <th>essay</th>\n",
       "      <th>price</th>\n",
       "      <th>quantity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>8393</td>\n",
       "      <td>p205479</td>\n",
       "      <td>2bf07ba08945e5d8b2a3f269b2b3cfe5</td>\n",
       "      <td>CA</td>\n",
       "      <td>2016-04-27 00:27:36</td>\n",
       "      <td>Grades PreK-2</td>\n",
       "      <td>Engineering STEAM into the Primary Classroom</td>\n",
       "      <td>I have been fortunate enough to use the Fairy ...</td>\n",
       "      <td>My students come from a variety of backgrounds...</td>\n",
       "      <td>Each month I try to do several science or STEM...</td>\n",
       "      <td>It is challenging to develop high quality scie...</td>\n",
       "      <td>My students need STEM kits to learn critical s...</td>\n",
       "      <td>53</td>\n",
       "      <td>1</td>\n",
       "      <td>Math_Science</td>\n",
       "      <td>AppliedSciences Health_LifeScience</td>\n",
       "      <td>AppliedSciences Health_LifeScience</td>\n",
       "      <td>I have been fortunate enough to use the Fairy ...</td>\n",
       "      <td>725.05</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0       id                        teacher_id school_state  \\\n",
       "0        8393  p205479  2bf07ba08945e5d8b2a3f269b2b3cfe5           CA   \n",
       "\n",
       "                 Date project_grade_category  \\\n",
       "0 2016-04-27 00:27:36          Grades PreK-2   \n",
       "\n",
       "                                  project_title  \\\n",
       "0  Engineering STEAM into the Primary Classroom   \n",
       "\n",
       "                                     project_essay_1  \\\n",
       "0  I have been fortunate enough to use the Fairy ...   \n",
       "\n",
       "                                     project_essay_2  \\\n",
       "0  My students come from a variety of backgrounds...   \n",
       "\n",
       "                                     project_essay_3  \\\n",
       "0  Each month I try to do several science or STEM...   \n",
       "\n",
       "                                     project_essay_4  \\\n",
       "0  It is challenging to develop high quality scie...   \n",
       "\n",
       "                            project_resource_summary  \\\n",
       "0  My students need STEM kits to learn critical s...   \n",
       "\n",
       "   teacher_number_of_previously_posted_projects  project_is_approved  \\\n",
       "0                                            53                    1   \n",
       "\n",
       "  clean_categories                 clean_subcategories  \\\n",
       "0     Math_Science  AppliedSciences Health_LifeScience   \n",
       "\n",
       "                       teacher_prefix  \\\n",
       "0  AppliedSciences Health_LifeScience   \n",
       "\n",
       "                                               essay   price  quantity  \n",
       "0  I have been fortunate enough to use the Fairy ...  725.05       4.0  "
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "price_data = resource_data.groupby('id').agg({'price':'sum', 'quantity':'sum'}).reset_index()\n",
    "project_data = pd.merge(project_data, price_data, on='id', how='left')\n",
    "print (price_data[1:3])\n",
    "project_data.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 220
    },
    "colab_type": "code",
    "id": "Hj4p9cEnU8C4",
    "outputId": "3581c26c-6cd1-45cc-8f75-3681022dec75"
   },
   "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>id</th>\n",
       "      <th>teacher_id</th>\n",
       "      <th>school_state</th>\n",
       "      <th>Date</th>\n",
       "      <th>project_grade_category</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>clean_categories</th>\n",
       "      <th>clean_subcategories</th>\n",
       "      <th>teacher_prefix</th>\n",
       "      <th>essay</th>\n",
       "      <th>price</th>\n",
       "      <th>quantity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>8393</td>\n",
       "      <td>p205479</td>\n",
       "      <td>2bf07ba08945e5d8b2a3f269b2b3cfe5</td>\n",
       "      <td>CA</td>\n",
       "      <td>2016-04-27 00:27:36</td>\n",
       "      <td>Grades PreK-2</td>\n",
       "      <td>Engineering STEAM into the Primary Classroom</td>\n",
       "      <td>I have been fortunate enough to use the Fairy ...</td>\n",
       "      <td>My students come from a variety of backgrounds...</td>\n",
       "      <td>Each month I try to do several science or STEM...</td>\n",
       "      <td>It is challenging to develop high quality scie...</td>\n",
       "      <td>My students need STEM kits to learn critical s...</td>\n",
       "      <td>53</td>\n",
       "      <td>1</td>\n",
       "      <td>Math_Science</td>\n",
       "      <td>AppliedSciences Health_LifeScience</td>\n",
       "      <td>AppliedSciences Health_LifeScience</td>\n",
       "      <td>I have been fortunate enough to use the Fairy ...</td>\n",
       "      <td>725.05</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0       id                        teacher_id school_state  \\\n",
       "0        8393  p205479  2bf07ba08945e5d8b2a3f269b2b3cfe5           CA   \n",
       "\n",
       "                 Date project_grade_category  \\\n",
       "0 2016-04-27 00:27:36          Grades PreK-2   \n",
       "\n",
       "                                  project_title  \\\n",
       "0  Engineering STEAM into the Primary Classroom   \n",
       "\n",
       "                                     project_essay_1  \\\n",
       "0  I have been fortunate enough to use the Fairy ...   \n",
       "\n",
       "                                     project_essay_2  \\\n",
       "0  My students come from a variety of backgrounds...   \n",
       "\n",
       "                                     project_essay_3  \\\n",
       "0  Each month I try to do several science or STEM...   \n",
       "\n",
       "                                     project_essay_4  \\\n",
       "0  It is challenging to develop high quality scie...   \n",
       "\n",
       "                            project_resource_summary  \\\n",
       "0  My students need STEM kits to learn critical s...   \n",
       "\n",
       "   teacher_number_of_previously_posted_projects  project_is_approved  \\\n",
       "0                                            53                    1   \n",
       "\n",
       "  clean_categories                 clean_subcategories  \\\n",
       "0     Math_Science  AppliedSciences Health_LifeScience   \n",
       "\n",
       "                       teacher_prefix  \\\n",
       "0  AppliedSciences Health_LifeScience   \n",
       "\n",
       "                                               essay   price  quantity  \n",
       "0  I have been fortunate enough to use the Fairy ...  725.05       4.0  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Convert NaN value to mean of the column\n",
    "project_data.fillna(project_data.mean(), inplace=True)\n",
    "project_data.head(1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "rEOuX_qsU8C9"
   },
   "source": [
    "## Splitting data into Train and cross validation(or test): Stratified Sampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "7cLRAQ5hU8C9"
   },
   "outputs": [],
   "source": [
    "y = project_data['project_is_approved'].values\n",
    "X = project_data.drop(['project_is_approved'], axis=1)\n",
    "\n",
    "# train test split\n",
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, stratify=y)\n",
    "X_train, X_cv, y_train, y_cv = train_test_split(X_train, y_train, test_size=0.33, stratify=y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "U4B41zHpU8DC"
   },
   "outputs": [],
   "source": [
    "catogories_essay = list(project_data['essay'].values)\n",
    "# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039\n",
    "\n",
    "# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/\n",
    "# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string\n",
    "# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python\n",
    "cat_essay_list = []\n",
    "for i in catogories_essay:\n",
    "    temp = \"\"\n",
    "    # consider we have text like this \"Math & Science, Warmth, Care & Hunger\"\n",
    "    for j in i.split(','): # it will split it in three parts [\"Math & Science\", \"Warmth\", \"Care & Hunger\"]\n",
    "        if 'The' in j.split(): # this will split each of the catogory based on space \"Math & Science\"=> \"Math\",\"&\", \"Science\"\n",
    "            j=j.replace('The','') # if we have the words \"The\" we are going to replace it with ''(i.e removing 'The')\n",
    "        j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:\"Math & Science\"=>\"Math&Science\"\n",
    "        temp+=j.strip()+\" \" #\" abc \".strip() will return \"abc\", remove the trailing spaces\n",
    "        temp = temp.replace('&','_') # we are replacing the & value into \n",
    "    cat_essay_list.append(temp.strip())\n",
    "    \n",
    "project_data['clean_essay'] = cat_essay_list\n",
    "\n",
    "from collections import Counter\n",
    "my_counter = Counter()\n",
    "for word in project_data['clean_essay'].values:\n",
    "    my_counter.update(word.split())\n",
    "\n",
    "cat_essay_dict = dict(my_counter)\n",
    "sorted_cat_essay_dict = dict(sorted(cat_essay_dict.items(), key=lambda kv: kv[1]))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "CkykgMd6U8DH"
   },
   "outputs": [],
   "source": [
    "catogories_title = list(project_data['project_title'].values)\n",
    "# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039\n",
    "# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/\n",
    "# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string\n",
    "# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python\n",
    "project_title_list = []\n",
    "for i in catogories_title:\n",
    "    temp = \"\"\n",
    "    # consider we have text like this \"Math & Science, Warmth, Care & Hunger\"\n",
    "    for j in i.split(','): # it will split it in three parts [\"Math & Science\", \"Warmth\", \"Care & Hunger\"]\n",
    "        if 'The' in j.split(): # this will split each of the catogory based on space \"Math & Science\"=> \"Math\",\"&\", \"Science\"\n",
    "            j=j.replace('The','') # if we have the words \"The\" we are going to replace it with ''(i.e removing 'The')\n",
    "        j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:\"Math & Science\"=>\"Math&Science\"\n",
    "        temp+=j.strip()+\" \" #\" abc \".strip() will return \"abc\", remove the trailing spaces\n",
    "        temp = temp.replace('&','_') # we are replacing the & value into \n",
    "    project_title_list.append(temp.strip())\n",
    "\n",
    "#project_data.drop('project_title', axis=1, inplace=True)\n",
    "#project_data['project_title'] = project_title_list\n",
    "\n",
    "from collections import Counter\n",
    "my_counter = Counter()\n",
    "for word in project_data['project_title'].values:\n",
    "    my_counter.update(word.split())\n",
    "\n",
    "project_title_dict = dict(my_counter)\n",
    "sorted_project_title_dict = dict(sorted(project_title_dict.items(), key=lambda kv: kv[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "PF0nuT5FU8DL",
    "outputId": "83743712-3123-40b9-bf31-4f3f33ee3675"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 33667/33667 [00:19<00:00, 1749.43it/s]\n"
     ]
    }
   ],
   "source": [
    "# Combining all the above stundents \n",
    "from tqdm import tqdm\n",
    "X_train_preprocessed_essays = []\n",
    "# tqdm is for printing the status bar\n",
    "for sentance in tqdm(X_train['essay'].values):\n",
    "    sent = decontracted(sentance)\n",
    "    sent = sent.replace('\\\\r', ' ')\n",
    "    sent = sent.replace('\\\\\"', ' ')\n",
    "    sent = sent.replace('\\\\n', ' ')\n",
    "    sent = re.sub('[^A-Za-z0-9]+', ' ', sent)\n",
    "    # https://gist.github.com/sebleier/554280\n",
    "    sent = ' '.join(e for e in sent.split() if e not in stopwords)\n",
    "    X_train_preprocessed_essays.append(sent.lower().strip())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "1jnsc4duU8DP",
    "outputId": "bd053e88-3871-4858-fefa-057e9e1d2ece"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 24750/24750 [00:14<00:00, 1685.42it/s]\n"
     ]
    }
   ],
   "source": [
    "# Combining all the above stundents \n",
    "from tqdm import tqdm\n",
    "X_test_preprocessed_essays = []\n",
    "# tqdm is for printing the status bar\n",
    "for sentance in tqdm(X_test['essay'].values):\n",
    "    sent = decontracted(sentance)\n",
    "    sent = sent.replace('\\\\r', ' ')\n",
    "    sent = sent.replace('\\\\\"', ' ')\n",
    "    sent = sent.replace('\\\\n', ' ')\n",
    "    sent = re.sub('[^A-Za-z0-9]+', ' ', sent)\n",
    "    # https://gist.github.com/sebleier/554280\n",
    "    sent = ' '.join(e for e in sent.split() if e not in stopwords)\n",
    "    X_test_preprocessed_essays.append(sent.lower().strip())\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "7dMEQbT0U8DT",
    "outputId": "bd3975bc-225c-4adb-89c2-2b4392b5a986"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 16583/16583 [00:09<00:00, 1672.94it/s]\n"
     ]
    }
   ],
   "source": [
    "# Combining all the above stundents \n",
    "from tqdm import tqdm\n",
    "X_cv_preprocessed_essays = []\n",
    "# tqdm is for printing the status bar\n",
    "for sentance in tqdm(X_cv['essay'].values):\n",
    "    sent = decontracted(sentance)\n",
    "    sent = sent.replace('\\\\r', ' ')\n",
    "    sent = sent.replace('\\\\\"', ' ')\n",
    "    sent = sent.replace('\\\\n', ' ')\n",
    "    sent = re.sub('[^A-Za-z0-9]+', ' ', sent)\n",
    "    # https://gist.github.com/sebleier/554280\n",
    "    sent = ' '.join(e for e in sent.split() if e not in stopwords)\n",
    "    X_cv_preprocessed_essays.append(sent.lower().strip())\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0cDINHK2U8DW"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "wJj8hTVpU8DY"
   },
   "source": [
    "## Step 3: Standardize (normalize) the data scale to prep for Naive Bayes algorithm.\n",
    "Because the distance between pairs of points plays a critical part on the classification, it is necessary to normalize the data This will generate an array of values. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "_MRmjcyoU8DZ"
   },
   "source": [
    "### 1.4.1 Vectorizing Categorical data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "t5tOM578U8Da"
   },
   "source": [
    "- https://www.appliedaicourse.com/course/applied-ai-course-online/lessons/handling-categorical-and-numerical-features/"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "JAGpFxofU8Db"
   },
   "source": [
    "### Vectorization of clean_categories for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 87
    },
    "colab_type": "code",
    "id": "yFhw8vQEU8Db",
    "outputId": "f0bac27d-3fae-409b-b671-6dbaffd35f39"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['Warmth', 'Care_Hunger', 'History_Civics', 'Music_Arts', 'AppliedLearning', 'SpecialNeeds', 'Health_Sports', 'Math_Science', 'Literacy_Language']\n",
      "Shape of matrix X_train_categories_one_hot  after one hot encodig  (33667, 9)\n",
      "Shape of matrix X_test_categories_one_hot after one hot encodig  (24750, 9)\n",
      "Shape of matrix X_cv_categories_one_hot after one hot encodig  (16583, 9)\n"
     ]
    }
   ],
   "source": [
    "# we use count vectorizer to convert the values into one \n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "vectorizer_clean_cat = CountVectorizer(min_df=10,ngram_range=(1,4), max_features=100, vocabulary=list(sorted_cat_dict.keys()), lowercase=False, binary=True)\n",
    "X_train_categories_one_hot = vectorizer_clean_cat.transform(X_train['clean_categories'].values)\n",
    "X_test_categories_one_hot = vectorizer_clean_cat.transform(X_test['clean_categories'].values)\n",
    "X_cv_categories_one_hot = vectorizer_clean_cat.transform(X_cv['clean_categories'].values)\n",
    "print(vectorizer_clean_cat.get_feature_names())\n",
    "print(\"Shape of matrix X_train_categories_one_hot  after one hot encodig \",X_train_categories_one_hot.shape)\n",
    "print(\"Shape of matrix X_test_categories_one_hot after one hot encodig \",X_test_categories_one_hot.shape)\n",
    "print(\"Shape of matrix X_cv_categories_one_hot after one hot encodig \",X_cv_categories_one_hot.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Sivr5hi9U8De"
   },
   "source": [
    "### Vectorization of clean_subcategories for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 107
    },
    "colab_type": "code",
    "id": "wV-_5vpGU8Df",
    "outputId": "b3cd1eac-fdec-4d91-86a9-c03c6f850f96"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['Economics', 'CommunityService', 'FinancialLiteracy', 'ParentInvolvement', 'Civics_Government', 'Extracurricular', 'ForeignLanguages', 'NutritionEducation', 'Warmth', 'Care_Hunger', 'SocialSciences', 'PerformingArts', 'CharacterEducation', 'TeamSports', 'Other', 'College_CareerPrep', 'History_Geography', 'Music', 'Health_LifeScience', 'EarlyDevelopment', 'ESL', 'Gym_Fitness', 'EnvironmentalScience', 'VisualArts', 'Health_Wellness', 'AppliedSciences', 'SpecialNeeds', 'Literature_Writing', 'Mathematics', 'Literacy']\n",
      "Shape of matrix X_train_sub_categories_one_hot  after one hot encodig  (33667, 30)\n",
      "Shape of matrix X_test_sub_categories_one_hot after oneX_test_sub_categories_one_hot  hot encodig  (24750, 30)\n",
      "Shape of matrixX_cv_sub_categories_one_hot after one hot encodig  (16583, 30)\n"
     ]
    }
   ],
   "source": [
    "# we use count vectorizer to convert the values into one \n",
    "vectorizer_sub_cat = CountVectorizer(min_df=10,ngram_range=(1,4), max_features=100,vocabulary=list(sorted_sub_cat_dict.keys()), lowercase=False, binary=True)\n",
    "X_train_sub_categories_one_hot = vectorizer_sub_cat.transform(X_train['clean_subcategories'].values)\n",
    "X_test_sub_categories_one_hot = vectorizer_sub_cat.transform(X_test['clean_subcategories'].values)\n",
    "X_cv_sub_categories_one_hot = vectorizer_sub_cat.transform(X_cv['clean_subcategories'].values)\n",
    "\n",
    "\n",
    "features_tfidf.extend(vectorizer_sub_cat.get_feature_names())\n",
    "features_std.extend(vectorizer_sub_cat.get_feature_names())\n",
    "features_bow.extend(vectorizer_sub_cat.get_feature_names())\n",
    "\n",
    "print(vectorizer_sub_cat.get_feature_names())\n",
    "print(\"Shape of matrix X_train_sub_categories_one_hot  after one hot encodig \",X_train_sub_categories_one_hot.shape)\n",
    "print(\"Shape of matrix X_test_sub_categories_one_hot after oneX_test_sub_categories_one_hot  hot encodig \",X_test_sub_categories_one_hot.shape)\n",
    "print(\"Shape of matrixX_cv_sub_categories_one_hot after one hot encodig \",X_cv_sub_categories_one_hot.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "6HfJD1uVU8Dl"
   },
   "outputs": [],
   "source": [
    "# you can do the similar thing with state, teacher_prefix and project_grade_category also"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "8zI0Iq8iU8Dn"
   },
   "source": [
    "### 1.4.2 Vectorizing Text data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "uYDxidiSU8Do"
   },
   "outputs": [],
   "source": [
    "# stronging variables into pickle files python: http://www.jessicayung.com/how-to-use-pickle-to-save-and-load-variables-in-python/\n",
    "# make sure you have the glove_vectors file\n",
    "with open('C:\\\\VipinML\\\\InputData\\\\glove_vectors', 'rb') as f:\n",
    "    model = pickle.load(f)\n",
    "    glove_words =  set(model.keys())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "67A3UkycU8Dr"
   },
   "source": [
    "### Vectorization of preprocessed_essays for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "HRfDbhhBU8Ds",
    "outputId": "075d78fc-81bb-434f-8f0f-7898ee4b0bdb",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 33667/33667 [00:11<00:00, 3016.17it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "33667\n",
      "300\n"
     ]
    }
   ],
   "source": [
    "# average Word2Vec\n",
    "# compute average word2vec for each review.\n",
    "X_train_avg_w2v_vectors = []; # the avg-w2v for each sentence/review is stored in this list\n",
    "for sentence in tqdm(X_train_preprocessed_essays): # for each review/sentence\n",
    "    vector = np.zeros(300) # as word vectors are of zero length\n",
    "    cnt_words =0; # num of words with a valid vector in the sentence/review\n",
    "    for word in sentence.split(): # for each word in a review/sentence\n",
    "        if word in glove_words:\n",
    "            vector += model[word]\n",
    "            cnt_words += 1\n",
    "    if cnt_words != 0:\n",
    "        vector /= cnt_words\n",
    "    X_train_avg_w2v_vectors.append(vector)\n",
    "\n",
    "print(len(X_train_avg_w2v_vectors))\n",
    "print(len(X_train_avg_w2v_vectors[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "mK3bEIwFU8Dv",
    "outputId": "95cfd275-d285-4531-cc9d-ee94da8df236"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 24750/24750 [00:07<00:00, 3119.96it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "24750\n",
      "300\n"
     ]
    }
   ],
   "source": [
    "# average Word2Vec\n",
    "# compute average word2vec for each review.\n",
    "X_test_avg_w2v_vectors = []; # the avg-w2v for each sentence/review is stored in this list\n",
    "for sentence in tqdm(X_test_preprocessed_essays): # for each review/sentence\n",
    "    vector = np.zeros(300) # as word vectors are of zero length\n",
    "    cnt_words =0; # num of words with a valid vector in the sentence/review\n",
    "    for word in sentence.split(): # for each word in a review/sentence\n",
    "        if word in glove_words:\n",
    "            vector += model[word]\n",
    "            cnt_words += 1\n",
    "    if cnt_words != 0:\n",
    "        vector /= cnt_words\n",
    "    X_test_avg_w2v_vectors.append(vector)\n",
    "\n",
    "print(len(X_test_avg_w2v_vectors))\n",
    "print(len(X_test_avg_w2v_vectors[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "nPDNW5DVU8Dz",
    "outputId": "65f9a210-0ee6-4096-d7b4-21fe75cb450f"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 16583/16583 [00:05<00:00, 3091.16it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "16583\n",
      "300\n"
     ]
    }
   ],
   "source": [
    "# average Word2Vec\n",
    "# compute average word2vec for each review.\n",
    "X_cv_avg_w2v_vectors = []; # the avg-w2v for each sentence/review is stored in this list\n",
    "for sentence in tqdm(X_cv_preprocessed_essays): # for each review/sentence\n",
    "    vector = np.zeros(300) # as word vectors are of zero length\n",
    "    cnt_words =0; # num of words with a valid vector in the sentence/review\n",
    "    for word in sentence.split(): # for each word in a review/sentence\n",
    "        if word in glove_words:\n",
    "            vector += model[word]\n",
    "            cnt_words += 1\n",
    "    if cnt_words != 0:\n",
    "        vector /= cnt_words\n",
    "    X_cv_avg_w2v_vectors.append(vector)\n",
    "\n",
    "print(len(X_cv_avg_w2v_vectors))\n",
    "print(len(X_cv_avg_w2v_vectors[0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "B5hVdV2EU8D2"
   },
   "source": [
    "### Vectorization of teacher_prefix  for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "635CBo7KU8D3",
    "outputId": "93d74ae1-b5b5-4749-ed01-f979e718a0e8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of matrix after one hot encodig  (33667, 30)\n"
     ]
    }
   ],
   "source": [
    "# we use count vectorizer to convert the values into one hot encoded features\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "vectorizer_teacher_prefix = CountVectorizer(min_df=10,ngram_range=(1,4), vocabulary=list(sorted_teacher_dict.keys()),max_features=100, lowercase=False, binary=True)\n",
    "X_train_teacher_prefix_data = X_train['teacher_prefix']\n",
    "\n",
    "X_train_teacher_prefix_data.fillna(\"Mrs.\", inplace = True) \n",
    "\n",
    "teacher_prefix_notnull = X_train_teacher_prefix_data[pd.notnull(X_train_teacher_prefix_data)]\n",
    "\n",
    "vectorizer_teacher_prefix.fit(teacher_prefix_notnull.values)\n",
    "\n",
    "X_train_teacher_prefix_one_hot = vectorizer_teacher_prefix.transform(teacher_prefix_notnull.values)\n",
    "print(\"Shape of matrix after one hot encodig \",X_train_teacher_prefix_one_hot.shape)\n",
    "\n",
    "features_bow.extend(vectorizer_teacher_prefix.get_feature_names())\n",
    "features_tfidf.extend(vectorizer_teacher_prefix.get_feature_names())\n",
    "features_std.extend(vectorizer_teacher_prefix.get_feature_names())\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "OZKY1V4_U8D6",
    "outputId": "dbc7978d-afe4-43a4-e615-588be040aa25"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of matrix after one hot encodig  (24750, 30)\n"
     ]
    }
   ],
   "source": [
    "# we use count vectorizer to convert the values into one hot encoded features\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "vectorizer_test_teacher_prefix_data = CountVectorizer(min_df=10,ngram_range=(1,4), max_features=100,vocabulary=list(sorted_teacher_dict.keys()), lowercase=False, binary=True)\n",
    "X_test_teacher_prefix_data = X_test['teacher_prefix']\n",
    "X_test_teacher_prefix_data.fillna(\"Mrs.\", inplace = True) \n",
    "teacher_prefix_notnull = X_test_teacher_prefix_data[pd.notnull(X_test_teacher_prefix_data)]\n",
    "vectorizer_test_teacher_prefix_data.fit(teacher_prefix_notnull.values)\n",
    "X_test_teacher_prefix_one_hot = vectorizer_test_teacher_prefix_data.transform(teacher_prefix_notnull.values)\n",
    "print(\"Shape of matrix after one hot encodig \",X_test_teacher_prefix_one_hot.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 72
    },
    "colab_type": "code",
    "id": "jVlSXhLiU8D-",
    "outputId": "530c6104-9566-4a9e-8685-ee5bf9adeeca"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['Economics', 'CommunityService', 'FinancialLiteracy', 'ParentInvolvement', 'Civics_Government', 'Extracurricular', 'ForeignLanguages', 'NutritionEducation', 'Warmth', 'Care_Hunger', 'SocialSciences', 'PerformingArts', 'CharacterEducation', 'TeamSports', 'Other', 'College_CareerPrep', 'History_Geography', 'Music', 'Health_LifeScience', 'EarlyDevelopment', 'ESL', 'Gym_Fitness', 'EnvironmentalScience', 'VisualArts', 'Health_Wellness', 'AppliedSciences', 'SpecialNeeds', 'Literature_Writing', 'Mathematics', 'Literacy']\n",
      "Shape of matrix after one hot encodig  (16583, 30)\n"
     ]
    }
   ],
   "source": [
    "# we use count vectorizer to convert the values into one hot encoded features\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "vectorizer_cv_teacher_prefix = CountVectorizer(min_df=10,ngram_range=(1,4), max_features=100,vocabulary=list(sorted_teacher_dict.keys()), lowercase=False, binary=True)\n",
    "X_cv_teacher_prefix_data = X_cv['teacher_prefix']\n",
    "X_cv_teacher_prefix_data.fillna(\"Mrs.\", inplace = True) \n",
    "teacher_prefix_notnull = X_cv_teacher_prefix_data[pd.notnull(X_cv_teacher_prefix_data)]\n",
    "vectorizer_cv_teacher_prefix.fit(teacher_prefix_notnull.values)\n",
    "print(vectorizer_cv_teacher_prefix.get_feature_names())\n",
    "X_cv_teacher_prefix_one_hot = vectorizer_cv_teacher_prefix.transform(teacher_prefix_notnull.values)\n",
    "print(\"Shape of matrix after one hot encodig \",X_cv_teacher_prefix_one_hot.shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "2UMRsjsEU8EB"
   },
   "source": [
    "### Vectorization of price for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 237
    },
    "colab_type": "code",
    "id": "YkncDNwBU8ED",
    "outputId": "d90086e7-8bbf-43fd-e4f7-16eb91b4df9c"
   },
   "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>id</th>\n",
       "      <th>teacher_id</th>\n",
       "      <th>school_state</th>\n",
       "      <th>Date</th>\n",
       "      <th>project_grade_category</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>clean_categories</th>\n",
       "      <th>clean_subcategories</th>\n",
       "      <th>teacher_prefix</th>\n",
       "      <th>essay</th>\n",
       "      <th>price</th>\n",
       "      <th>quantity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>47190</td>\n",
       "      <td>136494</td>\n",
       "      <td>p192637</td>\n",
       "      <td>82641393e4b305d09482f80a0f5585b1</td>\n",
       "      <td>LA</td>\n",
       "      <td>2016-11-18 16:29:24</td>\n",
       "      <td>Grades PreK-2</td>\n",
       "      <td>Ready, Set, Google!</td>\n",
       "      <td>\\r\\nMy students are so excited about all the n...</td>\n",
       "      <td>My students will be able to expand their knowl...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>My students need need two new Chromebook lapto...</td>\n",
       "      <td>2</td>\n",
       "      <td>Literacy_Language Math_Science</td>\n",
       "      <td>Literature_Writing Mathematics</td>\n",
       "      <td>Literature_Writing Mathematics</td>\n",
       "      <td>\\r\\nMy students are so excited about all the n...</td>\n",
       "      <td>296.377662</td>\n",
       "      <td>18.524903</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Unnamed: 0       id                        teacher_id school_state  \\\n",
       "47190      136494  p192637  82641393e4b305d09482f80a0f5585b1           LA   \n",
       "\n",
       "                     Date project_grade_category        project_title  \\\n",
       "47190 2016-11-18 16:29:24          Grades PreK-2  Ready, Set, Google!   \n",
       "\n",
       "                                         project_essay_1  \\\n",
       "47190  \\r\\nMy students are so excited about all the n...   \n",
       "\n",
       "                                         project_essay_2 project_essay_3  \\\n",
       "47190  My students will be able to expand their knowl...             NaN   \n",
       "\n",
       "      project_essay_4                           project_resource_summary  \\\n",
       "47190             NaN  My students need need two new Chromebook lapto...   \n",
       "\n",
       "       teacher_number_of_previously_posted_projects  \\\n",
       "47190                                             2   \n",
       "\n",
       "                     clean_categories             clean_subcategories  \\\n",
       "47190  Literacy_Language Math_Science  Literature_Writing Mathematics   \n",
       "\n",
       "                       teacher_prefix  \\\n",
       "47190  Literature_Writing Mathematics   \n",
       "\n",
       "                                                   essay       price  \\\n",
       "47190  \\r\\nMy students are so excited about all the n...  296.377662   \n",
       "\n",
       "        quantity  \n",
       "47190  18.524903  "
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.head(1)\n",
    "X_test.head(1)\n",
    "X_cv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 105
    },
    "colab_type": "code",
    "id": "P03kAOWIU8EF",
    "outputId": "0c20d86c-a98b-407d-ca73-9f966e88ce85"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "After vectorizations\n",
      "(33667, 1) (33667,)\n",
      "(24750, 1) (24750,)\n",
      "(16583, 1) (16583,)\n",
      "====================================================================================================\n"
     ]
    }
   ],
   "source": [
    "from sklearn.preprocessing import Normalizer\n",
    "normalizer = Normalizer()\n",
    "# normalizer.fit(X_train['price'].values)\n",
    "# this will rise an error Expected 2D array, got 1D array instead: \n",
    "# array=[105.22 215.96  96.01 ... 368.98  80.53 709.67].\n",
    "# Reshape your data either using \n",
    "# array.reshape(-1, 1) if your data has a single feature \n",
    "# array.reshape(1, -1)  if it contains a single sample.\n",
    "\n",
    "#normalizer.fit(X_train['price'].values.reshape(-1,1))\n",
    "\n",
    "X_train_price_norm = normalizer.transform(X_train['price'].values.reshape(-1,1))\n",
    "X_test_price_norm = normalizer.transform(X_test['price'].values.reshape(-1,1))\n",
    "X_cv_price_norm = normalizer.transform(X_cv['price'].values.reshape(-1,1))\n",
    "\n",
    "print(\"After vectorizations\")\n",
    "print(X_train_price_norm.shape, y_train.shape)\n",
    "print(X_test_price_norm.shape, y_test.shape)\n",
    "print(X_cv_price_norm.shape, y_cv.shape)\n",
    "print(\"=\"*100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "_kIodqX7U8EH"
   },
   "source": [
    "## Bag of words of project_title for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "lrl_bELDU8EI",
    "outputId": "3e24afce-908a-4c67-a10d-f893a2f7617d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of matrix X_train_project_title_bow after one hot encodig  (33667, 35459)\n",
      "Shape of matrix X_test_project_title_bow after one hot encodig  (24750, 35459)\n",
      "Shape of matrix X_cv_project_title_bow after one hot encodig  (16583, 35459)\n"
     ]
    }
   ],
   "source": [
    "# PROJECT_TITLE BOW\n",
    "# We are considering only the words which appeared in at least 10 documents(rows or projects). \n",
    "vectorizer_bow_project_title = CountVectorizer(min_df=10,ngram_range=(1,4),vocabulary=list(sorted_project_title_dict.keys()))\n",
    "X_train_project_title_bow = vectorizer_bow_project_title.fit_transform(X_train['project_title'])\n",
    "X_test_project_title_bow = vectorizer_bow_project_title.transform(X_test['project_title'])\n",
    "X_cv_project_title_bow = vectorizer_bow_project_title.transform(X_cv['project_title'])\n",
    "print(\"Shape of matrix X_train_project_title_bow after one hot encodig \",X_train_project_title_bow .shape)\n",
    "print(\"Shape of matrix X_test_project_title_bow after one hot encodig \",X_test_project_title_bow .shape)\n",
    "print(\"Shape of matrix X_cv_project_title_bow after one hot encodig \",X_cv_project_title_bow .shape)\n",
    "\n",
    "features_bow.extend(vectorizer_bow_project_title.get_feature_names())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "vDucjqqgU8EJ"
   },
   "source": [
    "## Bag of words of preprocessed_essays for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "C3ogOCXzU8EK",
    "outputId": "7eae7cb9-ec10-4280-8c65-261ea42ccf1d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of matrix X_train_text_bow after one hot encodig  (33667, 10499)\n",
      "Shape of matrix X_test_text_bow after one hot encodig  (24750, 10499)\n",
      "Shape of matrix X_cv_text_bow after one hot encodig  (16583, 10499)\n"
     ]
    }
   ],
   "source": [
    "# We are considering only the words which appeared in at least 10 documents(rows or projects).\n",
    "vectorizer_bow_essays = CountVectorizer(min_df=10)\n",
    "X_train_text_bow = vectorizer_bow_essays.fit_transform(X_train_preprocessed_essays)\n",
    "X_test_text_bow = vectorizer_bow_essays.transform(X_test_preprocessed_essays)\n",
    "X_cv_text_bow = vectorizer_bow_essays.transform(X_cv_preprocessed_essays)\n",
    "\n",
    "features_bow.extend(vectorizer_bow_essays.get_feature_names())\n",
    "\n",
    "#print (vectorizer_bow_essays.get_feature_names())\n",
    "print(\"Shape of matrix X_train_text_bow after one hot encodig \",X_train_text_bow.shape)\n",
    "print(\"Shape of matrix X_test_text_bow after one hot encodig \",X_test_text_bow.shape)\n",
    "print(\"Shape of matrix X_cv_text_bow after one hot encodig \",X_cv_text_bow.shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "rQhPrFajU8EL"
   },
   "source": [
    "## TFIDF of preprocessed_essays for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Ci6KRqEwU8EM"
   },
   "outputs": [],
   "source": [
    "tfidf_model = TfidfVectorizer()\n",
    "tfidf_model.fit(X_train_preprocessed_essays)\n",
    "# we are converting a dictionary with word as a key, and the idf as a value\n",
    "dictionary = dict(zip(tfidf_model.get_feature_names(), list(tfidf_model.idf_)))\n",
    "X_train_tfidf_words = set(tfidf_model.get_feature_names())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "PBUXhlNxU8ER"
   },
   "source": [
    "## TFIDF of preprocessed_essays for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "peZfxIS_U8ES",
    "outputId": "3b789dc2-5795-47ea-93f8-43400263af4a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of matrix X_train_text_tfidf after one hot encodig  (33667, 35746)\n",
      "Shape of matrix X_test_text_tfidf after one hot encodig  (24750, 35746)\n",
      "Shape of matrix X_cv_text_tfidf after one hot encodig  (16583, 35746)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "vectorizer_tfidf_essays = TfidfVectorizer()\n",
    "X_train_text_tfidf = vectorizer_tfidf_essays.fit_transform(X_train_preprocessed_essays)\n",
    "X_test_text_tfidf = vectorizer_tfidf_essays.transform(X_test_preprocessed_essays)\n",
    "X_cv_text_tfidf = vectorizer_tfidf_essays.transform(X_cv_preprocessed_essays)\n",
    "print(\"Shape of matrix X_train_text_tfidf after one hot encodig \",X_train_text_tfidf.shape)\n",
    "print(\"Shape of matrix X_test_text_tfidf after one hot encodig \",X_test_text_tfidf.shape)\n",
    "print(\"Shape of matrix X_cv_text_tfidf after one hot encodig \",X_cv_text_tfidf.shape)\n",
    "\n",
    "features_tfidf.extend(vectorizer_tfidf_essays.get_feature_names())\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "N5-0KZL-U8EV"
   },
   "source": [
    "## TFIDF of Project Title for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "5i_htKvAU8EW",
    "outputId": "7aaf8b75-5934-4861-bf08-8d794badb0b3"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of matrix  X_train_project_title_tfidf after one hot encodig  (33667, 1666)\n",
      "Shape of matrix  X_test_project_title_tfidf after one hot encodig  (24750, 1666)\n",
      "Shape of matrix  X_cv_project_title_tfidf after one hot encodig  (16583, 1666)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "vectorizer_tfidf_project_title = TfidfVectorizer(min_df=10)\n",
    "X_train_project_title_tfidf = vectorizer_tfidf_project_title.fit_transform((X_train['project_title']))\n",
    "X_test_project_title_tfidf = vectorizer_tfidf_project_title.transform((X_test['project_title']))\n",
    "X_cv_project_title_tfidf = vectorizer_tfidf_project_title.transform((X_cv['project_title']))\n",
    "print(\"Shape of matrix  X_train_project_title_tfidf after one hot encodig \",X_train_project_title_tfidf.shape)\n",
    "print(\"Shape of matrix  X_test_project_title_tfidf after one hot encodig \",X_test_project_title_tfidf.shape)\n",
    "print(\"Shape of matrix  X_cv_project_title_tfidf after one hot encodig \",X_cv_project_title_tfidf.shape)\n",
    "\n",
    "features_tfidf.extend(vectorizer_tfidf_project_title.get_feature_names())\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Nl7sCl4OU8Ea"
   },
   "source": [
    "### TFIDF AVG W2V for Project Title for X_train,X_test, X_cv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "3lPV9vGLU8Ea",
    "outputId": "842f34f8-d7cf-4dc2-da7e-047a6fa637c5"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 33667/33667 [00:00<00:00, 95889.22it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "33667\n",
      "300\n"
     ]
    }
   ],
   "source": [
    "# average Word2Vec\n",
    "# compute average word2vec for each review.\n",
    "X_train_project_title_avg_w2v_vectors = []; # the avg-w2v for each sentence/review is stored in this list\n",
    "for sentence in tqdm(X_train['project_title']): # for each review/sentence\n",
    "    vector = np.zeros(300) # as word vectors are of zero length\n",
    "    cnt_words =0; # num of words with a valid vector in the sentence/review\n",
    "    for word in sentence.split(): # for each word in a review/sentence\n",
    "        if word in glove_words:\n",
    "            vector += model[word]\n",
    "            cnt_words += 1\n",
    "    if cnt_words != 0:\n",
    "        vector /= cnt_words\n",
    "    X_train_project_title_avg_w2v_vectors.append(vector)\n",
    "\n",
    "print(len(X_train_project_title_avg_w2v_vectors))\n",
    "print(len(X_train_project_title_avg_w2v_vectors[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "_mF_-aNpU8Ed",
    "outputId": "a7d17462-3d97-4c8d-82f4-251b076c7f8f"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 24750/24750 [00:00<00:00, 105139.03it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "24750\n",
      "300\n"
     ]
    }
   ],
   "source": [
    "# average Word2Vec\n",
    "# compute average word2vec for each review.\n",
    "X_test_project_title_avg_w2v_vectors = []; # the avg-w2v for each sentence/review is stored in this list\n",
    "for sentence in tqdm(X_test['project_title']): # for each review/sentence\n",
    "    vector = np.zeros(300) # as word vectors are of zero length\n",
    "    cnt_words =0; # num of words with a valid vector in the sentence/review\n",
    "    for word in sentence.split(): # for each word in a review/sentence\n",
    "        if word in glove_words:\n",
    "            vector += model[word]\n",
    "            cnt_words += 1\n",
    "    if cnt_words != 0:\n",
    "        vector /= cnt_words\n",
    "    X_test_project_title_avg_w2v_vectors.append(vector)\n",
    "\n",
    "print(len(X_test_project_title_avg_w2v_vectors))\n",
    "print(len(X_test_project_title_avg_w2v_vectors[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 70
    },
    "colab_type": "code",
    "id": "j8Dn7swuU8Eg",
    "outputId": "caf87520-54fc-4d96-8146-695238baa920"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 16583/16583 [00:00<00:00, 92890.21it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "16583\n",
      "300\n"
     ]
    }
   ],
   "source": [
    "# average Word2Vec\n",
    "# compute average word2vec for each review.\n",
    "X_cv_project_title_avg_w2v_vectors = []; # the avg-w2v for each sentence/review is stored in this list\n",
    "for sentence in tqdm(X_cv['project_title']): # for each review/sentence\n",
    "    vector = np.zeros(300) # as word vectors are of zero length\n",
    "    cnt_words =0; # num of words with a valid vector in the sentence/review\n",
    "    for word in sentence.split(): # for each word in a review/sentence\n",
    "        if word in glove_words:\n",
    "            vector += model[word]\n",
    "            cnt_words += 1\n",
    "    if cnt_words != 0:\n",
    "        vector /= cnt_words\n",
    "    X_cv_project_title_avg_w2v_vectors.append(vector)\n",
    "\n",
    "print(len(X_cv_project_title_avg_w2v_vectors))\n",
    "print(len(X_cv_project_title_avg_w2v_vectors[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 158
    },
    "colab_type": "code",
    "id": "rqrB5vp4U8Ei",
    "outputId": "9ee47a4b-62b4-483d-eadd-6cabcee96996"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(33667, 61) (33667,)\n",
      "(16583, 61) (16583,)\n",
      "(24750, 61) (24750,)\n",
      "====================================================================================================\n",
      "(33667, 37473) (33667,)\n",
      "(16583, 37473) (16583,)\n",
      "(24750, 37473) (24750,)\n",
      "====================================================================================================\n"
     ]
    }
   ],
   "source": [
    "# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039\n",
    "from scipy.sparse import hstack\n",
    "\n",
    "\n",
    "X_tr = hstack((X_train_price_norm,X_train_sub_categories_one_hot,X_train_teacher_prefix_one_hot)).tocsr()\n",
    "X_cr = hstack((X_cv_price_norm,X_cv_sub_categories_one_hot,X_cv_teacher_prefix_one_hot)).tocsr()\n",
    "X_te = hstack((X_test_price_norm,X_test_sub_categories_one_hot,X_test_teacher_prefix_one_hot)).tocsr()\n",
    "\n",
    "#print (X_train_price_norm)\n",
    "X_tr_bow = hstack((X_train_price_norm,X_train_sub_categories_one_hot,X_train_teacher_prefix_one_hot,X_train_text_bow,X_train_project_title_bow)).tocsr()\n",
    "X_cr_bow = hstack((X_cv_price_norm,X_cv_sub_categories_one_hot,X_cv_teacher_prefix_one_hot,X_cv_text_bow,X_cv_project_title_bow)).tocsr()\n",
    "X_te_bow = hstack((X_test_price_norm,X_test_sub_categories_one_hot,X_test_teacher_prefix_one_hot,X_test_text_bow,X_test_project_title_bow)).tocsr()\n",
    "\n",
    "\n",
    "#print (X_train_price_norm)\n",
    "#X_tr_bow = hstack((X_train_price_norm,,X_train_teacher_prefix_one_hot,X_train_text_bow,X_train_project_title_bow)).tocsr()\n",
    "#X_cr_bow = hstack((X_cv_price_norm,,X_cv_teacher_prefix_one_hot,X_cv_text_bow,X_cv_project_title_bow)).tocsr()\n",
    "#X_te_bow = hstack((X_test_price_norm,,X_test_teacher_prefix_one_hot,X_test_text_bow,X_test_project_title_bow)).tocsr()\n",
    "\n",
    "\n",
    "X_tr_tfidf = hstack((X_train_sub_categories_one_hot,X_train_teacher_prefix_one_hot,X_train_price_norm,X_train_project_title_tfidf,X_train_text_tfidf)).tocsr()\n",
    "X_cr_tfidf = hstack((X_cv_sub_categories_one_hot,X_cv_teacher_prefix_one_hot,X_cv_price_norm,X_cv_project_title_tfidf,X_cv_text_tfidf)).tocsr()\n",
    "X_te_tfidf = hstack((X_test_sub_categories_one_hot,X_test_teacher_prefix_one_hot,X_test_price_norm,X_test_project_title_tfidf,X_test_text_tfidf)).tocsr()\n",
    "\n",
    "X_tr_tfidf_w2v = hstack((X_train_sub_categories_one_hot,X_train_teacher_prefix_one_hot,X_train_price_norm,X_train_project_title_avg_w2v_vectors)).tocsr()\n",
    "X_cr_tfidf_w2v = hstack((X_cv_sub_categories_one_hot,X_cv_teacher_prefix_one_hot,X_cv_price_norm,X_cv_project_title_avg_w2v_vectors)).tocsr()\n",
    "X_te_tfidf_w2v = hstack((X_test_sub_categories_one_hot,X_test_teacher_prefix_one_hot,X_test_price_norm,X_test_project_title_avg_w2v_vectors)).tocsr()\n",
    "\n",
    "\n",
    "X_tr_avg_w2v = hstack((X_train_sub_categories_one_hot,X_train_teacher_prefix_one_hot,X_train_price_norm,X_train_project_title_avg_w2v_vectors)).tocsr()\n",
    "X_cr_avg_w2v = hstack((X_cv_sub_categories_one_hot,X_cv_teacher_prefix_one_hot,X_cv_price_norm,X_cv_project_title_avg_w2v_vectors)).tocsr()\n",
    "X_te_avg_w2v = hstack((X_test_sub_categories_one_hot,X_test_teacher_prefix_one_hot,X_test_price_norm,X_test_project_title_avg_w2v_vectors)).tocsr()\n",
    "\n",
    "#print(\"Final Data matrix\")\n",
    "print(X_tr.shape, y_train.shape)\n",
    "print(X_cr.shape, y_cv.shape)\n",
    "print(X_te.shape, y_test.shape)\n",
    "\n",
    "print(\"=\"*100)\n",
    "\n",
    "print(X_tr_tfidf.shape, y_train.shape)\n",
    "print(X_cr_tfidf.shape, y_cv.shape)\n",
    "print(X_te_tfidf.shape, y_test.shape)\n",
    "print(\"=\"*100)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Kuc87Y4EU8El"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "9uLdC9Z7U8Eo"
   },
   "source": [
    "# Assignment 4: Naive Bayes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "O4V6GJ3IU8Ep"
   },
   "source": [
    "<ol>\n",
    "    <li><strong>Apply Multinomial NaiveBayes on these feature sets</strong>\n",
    "        <ul>\n",
    "            <li><font color='red'>Set 1</font>: categorical, numerical features + project_title(BOW) + preprocessed_eassay (BOW)</li>\n",
    "            <li><font color='red'>Set 2</font>: categorical, numerical features + project_title(TFIDF)+  preprocessed_eassay (TFIDF)</li>\n",
    "        </ul>\n",
    "    </li>\n",
    "    <br>\n",
    "    <li><strong>The hyper paramter tuning(find best Alpha)</strong>\n",
    "        <ul>\n",
    "    <li>Find the best hyper parameter which will give the maximum <a href='https://www.appliedaicourse.com/course/applied-ai-course-online/lessons/receiver-operating-characteristic-curve-roc-curve-and-auc-1/'>AUC</a> value</li>\n",
    "    <li>Consider a wide range of alpha values for hyperparameter tuning, start as low as 0.00001</li>\n",
    "    <li>Find the best hyper paramter using k-fold cross validation or simple cross validation data</li>\n",
    "    <li>Use gridsearch cv or randomsearch cv or you can also write your own for loops to do this task of hyperparameter tuning</li>\n",
    "        </ul>\n",
    "    </li>\n",
    "    <br>\n",
    "    <li><strong>Feature importance</strong>\n",
    "        <ul>\n",
    "    <li>Find the top 10 features of positive class and top 10 features of negative class for both feature sets <font color='red'>Set 1</font> and <font color='red'>Set 2</font> using values of `feature_log_prob_` parameter of  <a href='https://scikit-learn.org/stable/modules/generated/sklearn.naive_bayes.MultinomialNB.html'>MultinomialNB</a> and print their corresponding feature names</li>\n",
    "        </ul>\n",
    "    </li>\n",
    "    <br>\n",
    "    <li><strong>Representation of results</strong>\n",
    "        <ul>\n",
    "    <li>You need to plot the performance of model both on train data and cross validation data for each hyper parameter, like shown in the figure. Here on X-axis you will have alpha values, since they have a wide range, just to represent those alpha values on the graph, apply log function on those alpha values.\n",
    "    <img src='train_cv_auc.JPG' width=300px></li>\n",
    "    <li>Once after you found the best hyper parameter, you need to train your model with it, and find the AUC on test data and plot the ROC curve on both train and test.\n",
    "    <img src='train_test_auc.JPG' width=300px></li>\n",
    "    <li>Along with plotting ROC curve, you need to print the <a href='https://www.appliedaicourse.com/course/applied-ai-course-online/lessons/confusion-matrix-tpr-fpr-fnr-tnr-1/'>confusion matrix</a> with predicted and original labels of test data points. Please visualize your confusion matrices using <a href='https://seaborn.pydata.org/generated/seaborn.heatmap.html'>seaborn heatmaps.\n",
    "    <img src='confusion_matrix.png' width=300px></li>\n",
    "        </ul>\n",
    "    </li>\n",
    "    <br>\n",
    "    <li><strong>Conclusion</strong>\n",
    "        <ul>\n",
    "    <li>You need to summarize the results at the end of the notebook, summarize it in the table format. To print out a table please refer to this prettytable library<a href='http://zetcode.com/python/prettytable/'>  link</a> \n",
    "        <img src='summary.JPG' width=400px>\n",
    "    </li>\n",
    "        </ul>\n",
    "</ol>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "9vdB7g17U8Ep"
   },
   "source": [
    "<h4><font color='red'>Note: Data Leakage</font></h4>\n",
    "\n",
    "1. There will be an issue of data-leakage if you vectorize the entire data and then split it into train/cv/test.\n",
    "2. To avoid the issue of data-leakag, make sure to split your data first and then vectorize it. \n",
    "3. While vectorizing your data, apply the method fit_transform() on you train data, and apply the method transform() on cv/test data.\n",
    "4. For more details please go through this <a href='https://soundcloud.com/applied-ai-course/leakage-bow-and-tfidf'>link.</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "28WJkbDYU8Er"
   },
   "source": [
    "<h1>2. Naive Bayes Validation </h1>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Iff2jNuqU8Ew"
   },
   "source": [
    "<h2>2.4 Appling NB() on different kind of featurization as mentioned in the instructions</h2>\n",
    "\n",
    "<br>Apply Naive Bayes on different kind of featurization as mentioned in the instructions\n",
    "<br> For Every model that you work on make sure you do the step 2 and step 3 of instrucations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "XwxVvZReU8Ew"
   },
   "outputs": [],
   "source": [
    "def batch_predict(clf, data):\n",
    "    # roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class\n",
    "    # not the predicted outputs\n",
    "    y_data_pred = []\n",
    "    tr_loop = data.shape[0] - data.shape[0]%1000\n",
    "    # consider you X_tr shape is 49041, then your tr_loop will be 49041 - 49041%1000 = 49000\n",
    "    # in this for loop we will iterate unti the last 1000 multiplier\n",
    "    for i in range(0, tr_loop, 1000):\n",
    "        y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])\n",
    "    # we will be predicting for the last data points\n",
    "    if data.shape[0]%1000 !=0:\n",
    "        y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])\n",
    "    \n",
    "    return y_data_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DvnJV4TXU8Ez"
   },
   "outputs": [],
   "source": [
    "# we are writing our own function for predict, with defined thresould\n",
    "# we will pick a threshold that will give the least fpr\n",
    "def find_best_threshold(threshould, fpr, tpr):\n",
    "    t = threshould[np.argmax(tpr*(1-fpr))]\n",
    "    # (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high\n",
    "    print(\"the maximum value of tpr*(1-fpr)\", max(tpr*(1-fpr)), \"for threshold\", np.round(t,3))\n",
    "    return t\n",
    "\n",
    "def predict_with_best_t(proba, threshould):\n",
    "    predictions = []\n",
    "    for i in proba:\n",
    "        if i>=threshould:\n",
    "            predictions.append(1)\n",
    "        else:\n",
    "            predictions.append(0)\n",
    "    return predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "m4fMPZk4U8E1"
   },
   "outputs": [],
   "source": [
    "def multinomialNB_validation_alpha_analysis(X_tr,y_train,X_te,y_test):\n",
    "    import matplotlib.pyplot as plt\n",
    "    from sklearn.naive_bayes import MultinomialNB\n",
    "    from sklearn.metrics import roc_auc_score\n",
    "    from sklearn.model_selection import GridSearchCV\n",
    "   \n",
    "    #alphas = np.logspace(-1, 100, num=10)\n",
    "    #alphas = [0.09,10,2,3,4,5,0.0009,0.0001, 0.0003, 0.0004, 0.0008, 0.001,0.01, \\\n",
    "            # 0.07,0.09, 0.1]\n",
    "    alphas = [ 0.0001, 0.001, 0.01, 0.1,1,10,100,1000]\n",
    "\n",
    "    nb = MultinomialNB()\n",
    "    train_auc = []\n",
    "    test_auc = []\n",
    "    grid = GridSearchCV(estimator=nb,param_grid=dict(alpha= alphas))\n",
    "\n",
    "    %time grid.fit(X_tr, y_train)\n",
    "    \n",
    "    print(\"Best grid score: %s\" % grid.best_score_)\n",
    "    print(\"Best grid estimator: %s\" % grid.best_estimator_.alpha)\n",
    "    \n",
    "    alpha =  grid.best_estimator_.alpha\n",
    "    \n",
    "    NB =   grid.best_estimator_\n",
    "    \n",
    "    print(\"Best alpha is after applying GridSearchCV: %s\" % alpha)\n",
    "\n",
    "    # 4. make class predictions for X_test_dtm\n",
    "    y_test_pred = grid.predict_proba(X_te)    # 4. make class predictions for X_test_dtm\n",
    "    y_train_pred = grid.predict_proba(X_tr)\n",
    "\n",
    "    # calculate accuracy of class predictions\n",
    "    from sklearn import metrics\n",
    "   # print (metrics.accuracy_score(y_test, y_test_pred[:, 1]))\n",
    "\n",
    "    #the ROC curve is plotted with TPR against the FPR where TPR is on y-axis and FPR is on the x-axis.\n",
    "    train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred[:, 1])\n",
    "    test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred[:, 1])\n",
    "\n",
    "    plt.plot(train_fpr, train_tpr, label=\"train AUC =\"+str(auc(train_fpr, train_tpr)))\n",
    "    plt.plot(test_fpr, test_tpr, label=\"Test AUC =\"+str(auc(test_fpr, test_tpr)))\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"FPR\")\n",
    "    plt.ylabel(\"TPR\")\n",
    "    plt.title(\"AUC PLOT\")\n",
    "    plt.grid()\n",
    "    plt.show()\n",
    "     \n",
    "    print(\"=\"*100)\n",
    "    from sklearn.metrics import confusion_matrix\n",
    "    best_t = find_best_threshold(tr_thresholds, train_fpr, train_tpr)\n",
    "    print(\"Train confusion matrix\")\n",
    "    print(confusion_matrix(y_train, predict_with_best_t(y_train_pred[:, 1], best_t)))\n",
    "    print(\"Test confusion matrix\")\n",
    "    print(confusion_matrix(y_test, predict_with_best_t(y_test_pred[:, 1], best_t)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "cRZMkojYU8E5"
   },
   "source": [
    "### How to speculate the performance of the model using ROC Curve?\n",
    "#### An excellent model has AUC near to the 1 which means it has good measure of separability. A poor model has AUC near to the 0 which means it has worst measure of separability. In fact it means it is reciprocating the result. It is predicting 0s as 1s and 1s as 0s. And when AUC is 0.5, it means model has no class separation capacity whatsoever"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "TpPEeWZNU8E6"
   },
   "outputs": [],
   "source": [
    "def NB_best_validation(X_tr,y_train,X_te, y_test,best_alpha,count_vect):\n",
    "    # https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve\n",
    "    from sklearn.metrics import roc_curve, auc\n",
    "    from sklearn.naive_bayes import MultinomialNB\n",
    "    \n",
    "    # use model MultinomialNB, pass best aplha found already in earlier analysis.\n",
    "    nb = MultinomialNB(alpha=best_alpha,class_prior=[0.5,0.5])\n",
    "    nb.fit(X_tr, y_train)\n",
    "    # roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class\n",
    "    # not the predicted outputs\n",
    "\n",
    "    y_train_pred = nb.predict_proba(X_tr)\n",
    "    y_test_pred = nb.predict_proba(X_te)\n",
    "\n",
    "    #The ROC curve is plotted with TPR against the FPR where TPR is on y-axis and FPR is on the x-axis.\n",
    "    train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred[:, 1])\n",
    "    test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred[:, 1])\n",
    "\n",
    "    # plot AUC curve. AUC curve should show best accuracy rate, since best aplha is used in the logic.\n",
    "    plt.plot(train_fpr, train_tpr, label=\"train AUC =\"+str(auc(train_fpr, train_tpr)))\n",
    "    plt.plot(test_fpr, test_tpr, label=\"Test AUC =\"+str(auc(test_fpr, test_tpr)))\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"FPR\")\n",
    "    plt.ylabel(\"TPR\")\n",
    "    plt.title(\"AUC PLOT\")\n",
    "    plt.grid()\n",
    "    plt.show\n",
    "    \n",
    "   \n",
    "    print(\"=\"*100)\n",
    "    from sklearn.metrics import confusion_matrix\n",
    "    best_t = find_best_threshold(tr_thresholds, train_fpr, train_tpr)\n",
    "    print(\"Train confusion matrix\")\n",
    "    print(confusion_matrix(y_train, predict_with_best_t(y_train_pred[:, 1], best_t)))\n",
    "    print(\"Test confusion matrix\")\n",
    "    print(confusion_matrix(y_test, predict_with_best_t(y_test_pred[:, 1], best_t)))\n",
    "    \n",
    "    neg_class_prob_sorted = nb.feature_log_prob_[0, :].argsort()\n",
    "    pos_class_prob_sorted = nb.feature_log_prob_[1, :].argsort()\n",
    "\n",
    "   \n",
    "    print(\"Top 10 positive features %s\" % np.take(count_vect, pos_class_prob_sorted[:10]))\n",
    "    print(\"=\"*100)\n",
    "    #print(\"Top 10 negative features %s\" % np.take(count_vect, neg_class_prob_sorted[:10]))\n",
    "    \n",
    "  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "sI3M6hF-U8E9"
   },
   "outputs": [],
   "source": [
    "def NBAnalysis_cross_validation(X_train,y_train,X_test,y_test):\n",
    "    from sklearn import model_selection\n",
    "    from mlxtend.plotting import plot_decision_regions\n",
    "    from sklearn.naive_bayes import MultinomialNB\n",
    "    from sklearn.model_selection import train_test_split\n",
    "    from sklearn.metrics import accuracy_score,roc_auc_score\n",
    "    # Import classification report and confusion matrix to evaluate predictions\n",
    "    from sklearn.metrics import classification_report, confusion_matrix\n",
    "    import math\n",
    "    \n",
    "    train_auc = []\n",
    "    test_auc = []\n",
    "    error_rate = []\n",
    "    best_accuracy=0.0001\n",
    "    \n",
    "    #Alphas = [0.0000001,0.00001,0.0001, 0.001, 0.01, 0.06,0.07,0.1, 0.2,0.3,0.4,0.5,0.9,1,2]\n",
    "    Alphas = np.logspace(-3, 3, 30)\n",
    "    print (Alphas)\n",
    "    best_accuracy=0.0001\n",
    "    tunedParam = -10000\n",
    "    LogAlphas =[]\n",
    "    \n",
    "    for i in Alphas:\n",
    "        if (i >0  and math.log(i) >0):\n",
    "            TunedParameter = math.log(i)\n",
    "            LogAlphas.append(TunedParameter)\n",
    "            model = MultinomialNB(alpha=i,class_prior=[0.5,0.5]) # Number of nearest neighbor is i.\n",
    "\n",
    "         # fitting the model on crossvalidation train\n",
    "            model.fit(X_train, y_train)\n",
    "\n",
    "            # predict the response on the crossvalidation train\n",
    "            y_train_pred = model.predict(X_train)  # predicting the value using cross validation data. \n",
    "\n",
    "            # predict the response on the crossvalidation test\n",
    "            y_test_pred = model.predict(X_test)  # predicting the value using cross validation data. \n",
    "\n",
    "\n",
    "            # evaluate CV accuracy\n",
    "            acc = accuracy_score(y_test, y_test_pred, normalize=True) * float(100)  # I get the accuracy score. \n",
    "            if acc > best_accuracy:\n",
    "                best_accuracy =acc\n",
    "                tunedParam = TunedParameter\n",
    "                \n",
    "                \n",
    "           # print('\\n Test Accuracy for Tuned Parameter = %s is %s' % (TunedParameter, acc))\n",
    "           # print(\"=========================================\")\n",
    "\n",
    "            # roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class\n",
    "            # not the predicted outputs        \n",
    "            train_auc.append(roc_auc_score(y_train,y_train_pred))\n",
    "            test_auc.append(roc_auc_score(y_test, y_test_pred))\n",
    "        else:\n",
    "            continue\n",
    "    \n",
    "    print('\\n Test Accuracy for Tuned Parameter = %s is %s' % (tunedParam, best_accuracy))\n",
    "    \n",
    "   # print (LogAlphas)\n",
    "   # print (train_auc)\n",
    "    \n",
    "    plt.plot(LogAlphas, train_auc, label='Train AUC')\n",
    "    plt.plot(LogAlphas, test_auc, label='Test AUC')\n",
    "\n",
    "    plt.scatter(LogAlphas, train_auc, label='Train AUC points')\n",
    "    plt.scatter(LogAlphas, test_auc, label='Test AUC points')\n",
    "\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Log (Alphas)\")\n",
    "    plt.ylabel(\"AUC\")\n",
    "    plt.title(\"ERROR PLOTS\")\n",
    "    plt.grid()\n",
    "    plt.show()\n",
    "       "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "iosOz-smU8FB"
   },
   "source": [
    "<h2>2.4 Appling MultiNomialNB on different kind of featurization as mentioned in the instructions</h2>\n",
    "\n",
    "<br>Apply KMultiNomialNBNN on different kind of featurization as mentioned in the instructions\n",
    "<br> For Every model that you work on make sure you do the step 2 and step 3 of instructions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "FbniqbvcU8FB"
   },
   "source": [
    "<h2>2.3 Make Data Model Ready: encoding eassay, and project_title</h2>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "6DxjXFKoU8FC"
   },
   "source": [
    "### 2.4.0  Alpha Validation  for MultinomialNB on   Set 1: categorical, numerical features + project_title(BOW) + preprocessed_essay (BOW),<font color='red'> SET 1</font>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "OGHD6xEVU8FD"
   },
   "source": [
    "###  Alpha Validation  for MultinomialNB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 953
    },
    "colab_type": "code",
    "id": "lEXjcFcgU8FE",
    "outputId": "1276606b-c9aa-4af4-cfa7-fcf9c9076c47"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 167 ms\n",
      "Best grid score: 0.8484272432946208\n",
      "Best grid estimator: 1000\n",
      "Best alpha is after applying GridSearchCV: 1000\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "====================================================================================================\n",
      "the maximum value of tpr*(1-fpr) 0.27497395550552334 for threshold 0.961\n",
      "Train confusion matrix\n",
      "[[ 2719  2384]\n",
      " [13823 14741]]\n",
      "Test confusion matrix\n",
      "[[1295 1219]\n",
      " [6785 7284]]\n"
     ]
    }
   ],
   "source": [
    "multinomialNB_validation_alpha_analysis(X_tr,y_train,X_cr,y_cv)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Ej0C09rcU8FI"
   },
   "source": [
    "### 2.4.1  Alpha Validation  for MultinomialNB on  BOW - Set 1: categorical, numerical features + project_title(BOW) + preprocessed_essay (BOW),<font color='red'> SET 1</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "BagSSuAhU8FJ"
   },
   "outputs": [],
   "source": [
    "# Please write all the code with proper documentation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "ioZ8czwmU8FL"
   },
   "source": [
    "### kbest for BoW"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 953
    },
    "colab_type": "code",
    "id": "Fe_lmLbIU8FM",
    "outputId": "97a58e5a-24b9-4582-f129-f3b12d929e05"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 1.7 s\n",
      "Best grid score: 0.8484272432946208\n",
      "Best grid estimator: 10\n",
      "Best alpha is after applying GridSearchCV: 10\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "====================================================================================================\n",
      "the maximum value of tpr*(1-fpr) 0.0934538727668645 for threshold 1.0\n",
      "Train confusion matrix\n",
      "[[  488  4615]\n",
      " [  650 27914]]\n",
      "Test confusion matrix\n",
      "[[  166  2348]\n",
      " [  347 13722]]\n"
     ]
    }
   ],
   "source": [
    "multinomialNB_validation_alpha_analysis (X_tr_bow,y_train,X_cr_bow,y_cv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 953
    },
    "colab_type": "code",
    "id": "mNKo53m0U8FP",
    "outputId": "15315944-49e3-4ae5-b5c8-c035df03fc4f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 1.52 s\n",
      "Best grid score: 0.8484272432946208\n",
      "Best grid estimator: 10\n",
      "Best alpha is after applying GridSearchCV: 10\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "====================================================================================================\n",
      "the maximum value of tpr*(1-fpr) 0.0934538727668645 for threshold 1.0\n",
      "Train confusion matrix\n",
      "[[  488  4615]\n",
      " [  650 27914]]\n",
      "Test confusion matrix\n",
      "[[  166  2348]\n",
      " [  347 13722]]\n"
     ]
    }
   ],
   "source": [
    "\n",
    "multinomialNB_validation_alpha_analysis(X_tr_bow,y_train,X_cr_bow,y_cv)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "L-X5RIXTU8FR"
   },
   "source": [
    "### 2.4.2 Alpha Validation  for MultinomialNB on TFIDF Set 2: categorical, numerical features + project_title(TFIDF)+ preprocessed_essay (TFIDF),<font color='red'> SET 2</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 953
    },
    "colab_type": "code",
    "id": "GbJnvDI7U8FR",
    "outputId": "cdec24b2-aa83-4a02-d4b6-77a9c4eb241c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 1.71 s\n",
      "Best grid score: 0.8484272432946208\n",
      "Best grid estimator: 10\n",
      "Best alpha is after applying GridSearchCV: 10\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "====================================================================================================\n",
      "the maximum value of tpr*(1-fpr) 0.30281563192712685 for threshold 1.0\n",
      "Train confusion matrix\n",
      "[[ 2728  2375]\n",
      " [12384 16180]]\n",
      "Test confusion matrix\n",
      "[[1284 1230]\n",
      " [5912 8157]]\n"
     ]
    }
   ],
   "source": [
    "multinomialNB_validation_alpha_analysis(X_tr_tfidf,y_train,X_cr_tfidf,y_cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "XMZhiZcQU8FV"
   },
   "source": [
    "## 2.5.1 <font color='red'> Alpha-Analysis  <font color='blue'> categorical, numerical features + project_title(BOW) + preprocessed_essay (BOW),<font color='red'> SET 1</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 823
    },
    "colab_type": "code",
    "id": "gdDPJGVPU8FV",
    "outputId": "f3d9f039-b2ab-428d-c7b5-0b6a33598acc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.00000000e-03 1.61026203e-03 2.59294380e-03 4.17531894e-03\n",
      " 6.72335754e-03 1.08263673e-02 1.74332882e-02 2.80721620e-02\n",
      " 4.52035366e-02 7.27895384e-02 1.17210230e-01 1.88739182e-01\n",
      " 3.03919538e-01 4.89390092e-01 7.88046282e-01 1.26896100e+00\n",
      " 2.04335972e+00 3.29034456e+00 5.29831691e+00 8.53167852e+00\n",
      " 1.37382380e+01 2.21221629e+01 3.56224789e+01 5.73615251e+01\n",
      " 9.23670857e+01 1.48735211e+02 2.39502662e+02 3.85662042e+02\n",
      " 6.21016942e+02 1.00000000e+03]\n",
      "\n",
      " Test Accuracy for Tuned Parameter = 1.66738920527155 is 84.8398962793222\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "NBAnalysis_cross_validation(X_tr_bow,y_train,X_cr_bow,y_cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "ecObAWXmU8FW"
   },
   "source": [
    "### 2.5.2 <font color='red'> Alpha-Analysis   <font color='blue'>  on TFIDF Set 2: categorical, numerical features + project_title(TFIDF)+ preprocessed_essay (TFIDF),<font color='red'> SET 2</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 823
    },
    "colab_type": "code",
    "id": "r4NhWWncU8FX",
    "outputId": "5b5ef64e-79e8-493e-8451-eb747d59cc27"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.00000000e-03 1.61026203e-03 2.59294380e-03 4.17531894e-03\n",
      " 6.72335754e-03 1.08263673e-02 1.74332882e-02 2.80721620e-02\n",
      " 4.52035366e-02 7.27895384e-02 1.17210230e-01 1.88739182e-01\n",
      " 3.03919538e-01 4.89390092e-01 7.88046282e-01 1.26896100e+00\n",
      " 2.04335972e+00 3.29034456e+00 5.29831691e+00 8.53167852e+00\n",
      " 1.37382380e+01 2.21221629e+01 3.56224789e+01 5.73615251e+01\n",
      " 9.23670857e+01 1.48735211e+02 2.39502662e+02 3.85662042e+02\n",
      " 6.21016942e+02 1.00000000e+03]\n",
      "\n",
      " Test Accuracy for Tuned Parameter = 1.1909922894796787 is 84.8398962793222\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "NBAnalysis_cross_validation(X_tr_tfidf,y_train,X_cr_tfidf,y_cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "QCGuXvfLU8Fb"
   },
   "source": [
    "### 3.1 <font color='red'> ALPHA-BEST  <font color='blue'> categorical, numerical features + project_title(BOW) + preprocessed_essay (BOW),<font color='red'> SET 1</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 401
    },
    "colab_type": "code",
    "id": "5vRlhY20U8Fc",
    "outputId": "4d34095e-f6bf-40cd-a6d1-c48488283e1b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "====================================================================================================\n",
      "the maximum value of tpr*(1-fpr) 0.5680164428485288 for threshold 0.679\n",
      "Train confusion matrix\n",
      "[[ 3899  1204]\n",
      " [ 7329 21235]]\n",
      "Test confusion matrix\n",
      "[[1482 1032]\n",
      " [4115 9954]]\n",
      "Top 10 positive features ['mathematics' 'Glaze' 'Animals,' 'Apply' 'Micro' 'Dedicated' 'Parachutes'\n",
      " 'within' 'ENGAGED' 'Friends,']\n",
      "====================================================================================================\n",
      "Top 10 negative features ['mathematics' 'Exposure' 'Containers' 'board' 'Beginnings!' 'Discs'\n",
      " 'Carpet,' 'CPR' 'equal' 'Pay']\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Best alpha is found from above analysis of BoW. using best Alpha, test accuracy is calcuated as around.\n",
    "Best_alpha=0.1\n",
    "NB_best_validation(X_tr_bow,y_train,X_cr_bow, y_cv,Best_alpha,features_bow)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "bpfIpOEIU8Fe"
   },
   "source": [
    "### 3.2 <font color='red'> Alpha-Best   <font color='blue'>  on TFIDF Set 2: categorical, numerical features + project_title(TFIDF)+ preprocessed_essay (TFIDF),<font color='red'> SET 2</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 418
    },
    "colab_type": "code",
    "id": "oXWy1bWyU8Ff",
    "outputId": "7e9cdfce-c3bc-4a33-ea54-5314caa0cc4f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "====================================================================================================\n",
      "the maximum value of tpr*(1-fpr) 0.6641549848228029 for threshold 0.498\n",
      "Train confusion matrix\n",
      "[[ 4135   968]\n",
      " [ 5152 23412]]\n",
      "Test confusion matrix\n",
      "[[ 1147  1367]\n",
      " [ 3347 10722]]\n",
      "Top 10 positive features ['doormats' 'heaters' 'heater' 'reginald' 'wanting' 'heartrate' 'burets'\n",
      " 'burgeoning' 'sanitization' 'playspace']\n",
      "====================================================================================================\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nOzdeVxV1fr48c9iEhFwQEUFFecZHHDKUmxUmzQtNa9WVn67zVm3TBvv7f6absOtbLCyyXKorCzNyqtopSYOOOKAioojoIiAjOf5/bHhAAKCyOEA53m/XufFnvez4HCes/daey0jIiillHJdbs4OQCmllHNpIlBKKReniUAppVycJgKllHJxmgiUUsrFaSJQSikXp4lAKaVcnCYCVesZYyKNMaeMMXVKWH7XOcsijDHxheaNMeZBY8w2Y0yaMSbeGPO1MabHec6VYYxJNcYkGmMWGmOa56371BjzQin7GWPMP4wxe4wxZ40xB40xL+XHbIz5Oe+YqcaYbGNMVqH59y/2d6RcmyYCVasZY0KAywABbqjAIf4LPAQ8CDQCOgLfA9eeZ5/7RcQ3b9sGwBvlOM9bwBRgEuAHDAcuBxYAiMhwEfHNO+6XwCv58yJyTwXKpZSdh7MDUMrBJgFrgb+A24Cvy7ujMaYDcB8wUETWFVr1ZXn2F5GTxphvgb+X4zz3nnOe7caY0UCsMeZyEVle3riVulB6RaBqu0lYH9xfAtcYYwIvYN8rgPhzkkC5GWMaA6OBTRU5j4gcwkpiV1Xk/EqVlyYCVWsZYy4FWgMLRGQDsBe49QIOEQAcrcCp3zLGJAOb8/afWsb2jc9znqN565VyGE0Eqja7DfhVRBLz5r/KW5YvB/A8Zx9PIDtvOgloXoHzPigiDUQkSEQmiEhCGdsnnuc8zfPWK+UwmghUrWSMqQvcAgwxxhwzxhwDHgHCjDFheZsdBELO2bUNcCBv+n9AsDEm3MHhLgdaGmP6FV5ojGkJDMiLQymH0USgaquRQC7QFeiZ9+oC/I5VbwAwH7jDGNMvr/lmR6xkMQ9ARPYA7wJz85qVehljvI0x44wx0yoYl3veMfJfXiKyG3gf+NIYM8AY426M6QZ8CywTkWUVPJdS5aKJQNVWtwGfiMhBETmW/wLeASYYYzxE5BdgGvAJcBpYAnwGzCp0nAfz9pkJJGPVM4wCfqxgXNOAs4Ve+a2B7gc+AuYAqcBSIBKrslkphzI6MI1SSrk2vSJQSikXp4lAKaVcnCYCpZRycZoIlFLKxdW4voYaN24sISEhFdo3LS2NevXqVW5A1ZyW2TVomV3DxZR5w4YNiSLSpKR1NS4RhISEsH79+grtGxkZSUREROUGVM1pmV2Dltk1XEyZjTEHSlunt4aUUsrFaSJQSikXp4lAKaVcnCYCpZRycZoIlFLKxTksERhjZhtjThhjtpWy3hhj3jLGxBpjthhjejsqFqWUUqVz5BXBp8Cw86wfDnTIe00B3nNgLEoppUrhsOcIRGSVMSbkPJvcCHwuVvena40xDYwxzUWkIkMDKqVqERHBJpCZk0tWjo2sXBtZOTbSs3IRgexcG/tO5+J34BS5NiE1MxuDAWPtm2sDmwg2m3Ucm4j9lWuDpNRM/Lw9EQQREOuk2MTaX6xZbHm9M1vb5K+nYD8p2F8EzmRkYxPwreNeab8Lj9wMQpLX0DRtN8n1ugARlXZs+zkq/YjlFwQcKjQfn7esWCIwxkzBumogMDCQyMjICp0wNTW1wvvWVFpm1+CIMosIuQIZOZBtE7JtkJ4t5NjgTN5Pm0COTUjJsrbHQI4Ncm1wMkPwcodsGxxLs1HP05Btg8wcIeGs4O0BCemCpxsYYx0rV8j74C5nkGtWV2qZK5OpwD5eZNPd7Ocq9w30cdtFiDlOE3Pavv77+rc55L3tzERQ0u+pxD+/iMwib7CQ8PBwqeiTdfokomvQMhcQEU6fzSY5PZsTZzKJP5XOgaR03N0MaVk5HEhMx8fLnTOZOaSczWb38TPU8XDnWErGRcdkjPUtuYlfHdyNB8k5QvMGdWno7oZ/faGOhxuXNqhLcnoWLRv54OFmcHdzw8PN4OZm8HAzJKdnE9ywLl4ebni5u5GelUNTf2/c3Qwx27fRq2dY3n6G7Fwb9ep44G4Mbsbg5gZuxlrnZsAYU2QdgKe7m/VBZKxtDdZ2BmseY5Ujf96azvtZeLrwenMBKSAnCzZ+BtFfwpFNxde714HA3tA2AvreSYNNsQ55bzszEcQDLQvNBwNHnBSLUjXC2axcEs5kcjj5LNm5Ng4nn+XY6Qz2x2WxOGEzx1IySMvMIfZEKikZOWUez8PNUMfDjeCGPtSv60lYywZkZOcS0akJnu5uNKvvTXaujfp1PfH39sQY8K3jQQMfL9zdDP7eHtaHdN4HtbenO57ubni6mwv7QKyAOgk7GdKxxK5zqq/0k7BrCRzeaCUA2zl/o9aXQpvBEHIphAwq4QCxDgnLmYlgEXC/MWYe0B84rfUDypVl5dg4npJB9KFk4hLTSMnIZuPBZJJSM0nPyiUlI5uMbFup+zf2PUFQQx/86njQr00AIkKjel4ENaxLoL83zet7E9ywLg19vPD19sDL3c3hH9YKOLoZ/voAti2EnLNF1zXvCd1vgm43QYOWJe9fBRyWCIwxc7FqNRobY+KBZwFPABF5H2t82BFYKS4duMNRsShVXSSlZnIsJYODSensOn6Gg0npbDqUTGZ2LkdOF78d06qRD8ZAm8b16NaiPo39vAio54WPlwfN6nvj4+VOY986bN+whsuHDnVCiVQxKUcg5idYNwuS9hRd16AVDLgPuo0Cv0DnxFcCR7YaGl/GegHuc9T5lXKGrBwbB0+mcTg5g4Mn09lyKJk9J1LJyrFu45w+m11k+xb1vWnRoC6e7m6M6RNMXS8Pugf506ZxPQL9vfF0L18L7xj9Zu882Wfhy5utBHByb/H1XW6AvndB60vA3bPq4yuHGtcNtVLVQUZ2LlsPn+bwqbPEHE1h08FkYo6lkJqZg5TQ5KFDU1+uC21Oy0Y+NPTxpH1TP7o098PHS/8FaxwRSE+CzfNgx/cQH1WwrtO1ULchtLkMOl5jTdcA+i5U6jxOpmWx4cAptsYnk5SWReSuBA4nF73Pawy0CahHr1YNae7vTZ/WDWkd4EPLRj4E5rVwUTVcdgas/9j68D+2peg6z3rQczyM+I/1ZqiBNBEolyciJKdnczj5LNGHktmfmMbhU2dZf+AUiamZRbYd0LYRXVv4E9SgLiEBPvRp3YgOgb54e1beA0TKyWy5cHgDbPoC4tfDiR1F19drCl1vhMBu0Pk68K1hLZdKoIlAuZSsHBvrD5zkQFI6v+9J4OjpDPYnppGcXvTevYebYVj3ZoQG16dTM386NPWlmb83bvrtvnay2WDtu7BpDiTEFF3XuBO0GgAB7aDXRPBp5JwYHUgTgarVRITtR1JYFnOc9XGn+CM20b7Ot44HAb5eRHRsQrcW9WnZqC5BDXwIaeyDbx0PbVpZW4nAgT+tVj3Jh+BUHJw9WbDewxv63W217GneE9xq/9WeJgJVa2Rk57JmXxLf7s5izoH1/LUvidSsgspbHy93Jg1sTZfm/vRs2YBOgX76Dd9V5GRZT+/++jRknSm6rmk3COoDzUPh0kegjp9zYnQiTQSqxsrKsbFmXxIrdp5g3f6T7DiaUmjtcdo39eWyQF8GtmtM35CGdAr002/5ruR0PCx9EmKXQXZ60XWDH4cu11sf/koTgao5RIRNh5L5c08im+NP80dsgv1J235tGjFlcFtCAupR52Qso4df7uRolVOciLHu9W/8vOhy32ZwyQNWJa8Tn+CtrjQRqGor1ybsOXHGure/J5Gl24/Z1zX2rcP1oS0Y3LEJgzs0ob5PwYM6kZH7nBGuqmLGlg27lsKhv6y2/HG/F92gaTcY/Ch0HekS9/kvhiYCVW2ICDuOpvBnbCIfrNxHUlqWfV1jXy9u7NmCTs38uLFnEEEN6joxUuUUtlzY8AnELof4KIaknSi63sMbWvaDIU9Aq0uwdzGqyqSJQDmNiLAl/jS/bD9GVNxJouJO2dcF+tehX0gjrujSlP5tAwgLrq/3912NCBxcA9u/s/ruOVOoc2J3L9J8gqkXNtLqwiG4r37wXwRNBKrKnTiTQeTOBN6NjCUuyarE83AzDGofQIemftzavxUdmvrqB7+rOpsMn98IR6OLLm/Uzuq24dJHwLcpUS447oSjaCJQVeJw8ll+3nqU1XuTWL7TuqQP9K/D9BGduaZbM1oH1HNyhMqpzp6C7+6xrgAyCkbkoucECL8TgnrX2O4bagJNBMph9hw/w6LNR4jclcDWwwX/3H8b0IprujXjknaNtR8eV5abDX/+F1a+DLkF9UG0GWxV8Pa5XSt5q4gmAlWpdh5L4eetx1izL4l1+62nNX283Hnoig5c0i6A7kH1qVdH33YuKyMF/vdPq6VP4c7b6reCAX+HflPAXd8fVU1/4+qiZefaeGd5LHPXHeTEmYJO2saGt+S2S0Lo3Eyf4HVpGSmw7Dmrbb+tUJ9O9Zpa9/wjnoT6QU4LT2kiUBdhxc4TvBsZy+b402Tl2Aj0r8MNYS34vyFt6dLMXz/8XV3cH/D5yKIf/o07Wt/6+9yh3/yrEf1LqAtiswnfbIhn9p/72XnsDN6eblzSLoBb+7Xiqq6B2tLH1R1YAz8+CIm7iy6/4R3oeave86+mNBGoclm1O4Hf9ySwZOsx+8AsD1/ZgXuGtNO++F3d1m/gx4cgK7Xo8g7XwJDHITjcOXGpctNEoEqVmpnDh6v2sWTrUfacsP7Jw4Lrc9/Q9ozt21Jb/Li6U3Hw37Ciywbca3Xf3LKfU0JSFaOJQJXoi7UHePaHbdjyunC+Z0g77r+8Pb7a4se12XKtcXpXvgIJO61lxh3u/h+06OXc2FSF6X+1sktMzWTO2gOs2JXA5kPJNKrnxYwRXRjZK0i//bu6fSvhp4fhZKEO/Yw7XPG09aSvqtE0EShycm28sDiGT1fHAVZ3Dw9e3p47L21bpFdP5WKObYPfnoa9y4sub38lXPcGNGjlnLhUpdNE4MJOn83mvci9zP5jP1m5Nvy8PXjn1t5c2l6f+HVph6Jg3q1QuHfPdpdbH/4NQ5wWlnIcTQQuJifXxvoDp3j1l11sOngKm0D/No0Y27clI3sGadt/V/bLDFjzTtFlE7+HdkOdE4+qMpoIXISIsGTrMe77aiMAnu6GPq0b8tAVHRnUPkDb/7simw22L7T6+inc7r9RW7j2NesqQLkETQS1XLZN+Oqvg8xcEcvh5LO4GRjbtxXThnemfl29/++y0hLh1XZFl4XdClc9D75NnROTchpNBLXYbzuOc89v6eTKVpr61eGfN3ZjXN9WeHnoAB4u7fAG+LDQt/1Hd4FfM+fFo5xOE0EtFHM0hcmfRnH0dAYA/7imE3df1lYTgCsTgQWTIGZRwbImXeC+tc6LSVUbmghqkYzsXJ76fhvfbIgH4JbwYIbUP8m1Q9s7OTLlNKcOWOP8/vFGwbJWl8DA+6DLdc6LS1UrmghqieMpGdz07moOJ58lNLg+z17flT6tGxEZGens0FRVE7E++P/3fNHldRvBI9vBy8c5calqy6GJwBgzDPgv4A58JCIvnbO+FfAZ0CBvm2kissSRMdVGy3ce5/6vNpGelcttA1vz/I3dnR2SqmJuuZlWtw+b58HJvQUrPOrCzZ9aD4Fpt8+qFA57Zxhj3IGZwFVAPBBljFkkIjsKbfYUsEBE3jPGdAWWACGOiqm2Sc3MYfrCrSzafAR3N8NXd/fnknaNnR2WqioisP07iPqIwQf+LFjuHwRdrofLn4I6fs6LT9UYjvyK0A+IFZF9AMaYecCNQOFEIIB/3nR94IgD46k1snNtPLdoO1/+dRCAm3oH8ciVHWnZSC/5XcaeZfDlaPuszXjg1utWGPEaeHg5MTBVExkRccyBjRkDDBORu/LmJwL9ReT+Qts0B34FGgL1gCtFZEMJx5oCTAEIDAzsM2/evArFlJqaiq+vb4X2rQ5EhIV7svlxnzXiU/N6hlHtvejXvPR8XtPLXBG1ucxuuZn03vgPfNMOAJDt4UdU37c4me1Va8tcmtr8dy7NxZR56NChG0SkxMEhHHlFUNKjqudmnfHApyLymjFmIPCFMaa7iNiK7CQyC5gFEB4eLhERERUKKDIykoru62zr407y3I/b2XY4Hb86HtzUO4jnbuhW5hPBNbnMFVVry/zVONj9c8H80Bl4DnmcS6jFZT4PLXPlcWQiiAdaFpoPpvitnzuBYQAissYY4w00Bk6g7GZ8t9V+G+i60Ob8d1wv7RTOVeTmQPQcawSwfP3vgWEvgXYLoiqJIxNBFNDBGNMGOAyMA249Z5uDwBXAp8aYLoA3kODAmGqUhDOZvL18D1/+dZDWAT4s/PslBPjWcXZYqirkZML6T2DpE0WXP74ffBo5JyZVazksEYhIjjHmfuAXrKahs0VkuzHmn8B6EVkEPAp8aIx5BOu20e3iqEqLGuRUWhZvLd/DJ3/GATC4YxNmTeyjYwPXdiLw7V1WSyDJLVjeohf8baEmAOUwDm1YnPdMwJJzlj1TaHoHMMiRMdQ0a/YmcednUaRn5RIS4MPT13Xlii6Bzg5LOdrpw/BG14L5Jp2h19+g9yTwru+8uJRL0CdMqpH/xRznzs/W09DHk/+O68VVXTUB1Go2G/zvOWsgmIOrC5Y/laBNQFWV0kRQDRw6mc7j32xhzb4kmvl789Xd/WnbxLWaxbmc318v2gVEvSZwyYMw6EHnxaRcliYCJztxJoPLXlkBwAOXt+fvEe3w8dI/S62UlQ4fDoWEnQXLmnaDe34HN63/Uc6jnzhOtOf4GW796C8AXhkTyi3hLcvYQ9VIGSmwYCLsiyxY1v5KGPUB1NMuQZTzaSJwkj9jE5mQlwQ+m9yPIR2bODkiVem+vh32LoeM0wXLBt4PV7+gzwCoakUTgRN89Ps+XlgcQ0MfTz66LZw+rbVZYK2RlmT1/7/8XwXLOl8HLftZSUBvAalqSBNBFfvo9338e0kMPl7u/PzQYJrV93Z2SKoyZKTApyPg2Naiy/+xV2//qGpPE0EVyc61cenLyzmekkm3Fv58eVd/GvhoE8Eab/Xb1hPAhccAuHQqXPKAPgCmagxNBFXkqtdXcjwlk4B6Xnz790v0KeGaLiMFXg4peALYrwWEjYMrn3VqWEpVhCaCKjBv3UHiktKp6+lO1IwrcdMO42qurHR4py+kxBcse2AjBLRzXkxKXSRNBA4WeyKVaQu34uFm+PWRwZoEarLoufD9PQXzV/3LugWkLYBUDaeJwMHu/dIaZ2fmhN46glhNJQIfXFZQEdyiN0xZ4dyYlKpEmggcaPOhZHYfT8XDzXBNt2bODkdVxJqZ8Mv0gvn/WwXNw5wXj1IOoInAQQ6dTOfGmdaA4vOmDHByNOqC/TQV1n9cMO9RF/6xRweDV7WSJgIHEBFum70OgEev6kh4iDYjrDG2fw9f31Yw3+UGuPEd7Qpa1WqaCBzg5aW72JeYxg1hLXjgig7ODkeVRQTWfQg//6NgWeOOMH6etgZSLkETQSUSEV5euov3V+6lUT0vXr9F7yVXa7ZcKwGcOxzk+PnQaZhzYlLKCTQRVKL3V+7j/ZV7adekHnPu6o+Hu5uzQ1KlORINs4YUXfbEAajbwDnxKOVEmggqSfShZF5eupMmfnVY/OBl+uRwdZV8CBZMgiMbrfnGnaymoF71nBuXUk6kiaASxCWmMXLmn3i6G+bePUCTQHX16XUQ93vB/JBpMPRJ58WjVDWhiaASPDQ/GoAPJvahfVMdYrJa+mAIHLX+Toz4D/S727nxKFWNaCK4CDabEP7vZZxMy2JQ+wAu76yDzVc72Wfhy5sLksDj+7VXUKXOoYngIgx9LZKTaVl0a+HPF5P7Ozscda5fn7K6ic5386eaBJQqgSaCCvrvsj0cSEoH4Pv7BmlnctXNxi8KksCA+6y6AH0qWKkSaSKogKTUTN5YthuArc9djac2E61WWh5cCPs+s2Zu/Ro6Xu3cgJSq5jQRVMC/l8QAMGtiH/y8PZ0cjbJLPwkz+9Mu7YQ1P/hxTQJKlYMmggt0Oj2bhRsP4+XuxtXao2j1kJ0BvzwJ62cXLLt7BQT1dl5MStUgmggu0Gu/7QLg3qHaB021sOMH6wGxfIE9iOzyAhGaBJQqN00EF2hz/Gl8vNx5+MqOzg7FtWWfhfl/g9hl1nyrgXD7EnBzg8hIp4amVE2jieACiAibDyVzQ1gLZ4fi2jJT4cWggvlRsyBsrPPiUaqGc2giMMYMA/4LuAMfichLJWxzC/AcIMBmEbnVkTFdjI0HTwHg56350ykyUuD9QZB8sGDZs8k6ZrBSF8lhn2jGGHdgJnAVEA9EGWMWiciOQtt0AJ4EBonIKWNMU0fFUxn2JqQBcFPvoDK2VJVu/+/w2XUF85c9Bpc/pUlAqUrgyK+2/YBYEdkHYIyZB9wI7Ci0zd3ATBE5BSAiJxwYz0UREWZ8t5W6nu50D9LRqqrMhs/gxwcL5r384MlDmgCUqkRGRBxzYGPGAMNE5K68+YlAfxG5v9A23wO7gUFYt4+eE5GlJRxrCjAFIDAwsM+8efMqFFNqaiq+vhXrFG5lfDafbMvisiAP7uxRp0LHcIaLKbMzueekcdkfBXcJM+o0YVv3J0n1K7u1Vk0t88XQMruGiynz0KFDN4hIeEnrHHlFUNJXtnOzjgfQAYgAgoHfjTHdRSS5yE4is4BZAOHh4RIREVGhgCIjI6novjNeWg7Am3dEEOBbcxLBxZTZKWy58HZvOBVXsOy+dXg36USJ7+AS1LgyVwIts2twVJkdmQjigZaF5oOBIyVss1ZEsoH9xphdWIkhyoFxXbA/YxM5nHyWW/u3qlFJoMZZ8671YFi+4a9Avyl6G0gpB3NkIogCOhhj2gCHgXHAuS2CvgfGA58aYxoDHYF9DoypQv71k1WtMeWytk6OpJbKzYZFD8Dmuda8R12rHsBdu+9Qqio4LBGISI4x5n7gF6z7/7NFZLsx5p/AehFZlLfuamPMDiAX+IeIJDkqpopIzcxh57EzeHm4EdJYhzOsVDYbzBsPuwtVC13zIgy813kxKeWCHNogXkSWAEvOWfZMoWkBpua9qqVZK/cCMHFAaydHUsuIwD8bFsx3vg5Gva9dRSvlBPpkVBneWRELwNPXdXVyJLXIsa3w/qUF8zOOg6e38+JRysVpIjiP2BOp2AQaawVx5fnpkYJeQht3hL+v1roApZxME8F5LNwYD8Brt4Q5OZJaYvm/C5LAyPeh53jnxqOUAjQRnNfvexIBGNyhsZMjqeFE4D8dIX/AmOv/q0lAqWpEE8F5bD18moY+nhhtx15xNlvRSuGHNkPDEKeFo5QqThNBKXYdOwPA1V11FLIKS9wD7xR6HvjpRK0PUKoauuBR140x7saYCY4Ipjr5YJXVbPTKroFOjqSGWvte0STw1AlNAkpVU6VeERhj/IH7gCBgEfAbcD/wGBANfFkVATrLvrwup6/SRHBhdi6GeYUeIO83BUa86rx4lFJlOt+toS+AU8Aa4C7gH4AXcKOIRFdBbE5jswk7jqRwbWhzZ4dSs8QuK0gCgT1gxCvQ+hLnxqSUKtP5EkFbEekBYIz5CEgEWonImSqJzIlm/7mfrFwbA9o0cnYoNcfBv2DOaGt6+KvQf4pz41FKldv5EkF2/oSI5Bpj9rtCEgBYFnMcgBt76UhkZbLZYOFdsO1ba75+K00CStUw50sEYcaYFArGFahbaF5ExN/h0TmBzSas3XeSZv7e+Htr5WaZXgyC7HRrWh8SU6pGKjURiIh7VQZSXXy1zhoY/fowrR84r2PbrIHk8z1zCtwuuBGaUqoaOF+rIW/gHqA9sAWrG+mcqgrMWd5evgeAR6/u5ORIqrHM1IIk4BsI96/XJKBUDXa+/97PgHBgKzACeK1KInKy4ymZBDWoi7enS14QlS3yZet2EECjtvDYbvCulXcJlXIZ56sj6Fqo1dDHwLqqCcl5Vuyy+sIZ2rmJkyOppjbNgcj/Z033ud3qM0gpVeOVt9VQjiv0t/P/FscAcOelOiRlESLw9W2w4wdr/ro3IfwO58aklKo050sEPfNaCYHVUqhWtxrKybWx50QqQzs1oY0OSVnUx1dDfN4F4a0LoOM1zo1HKVWpzpcINotIryqLxMnWHzgFQP+2AU6OpJp5uw8kWaO08fh+8NGH7JSqbc6XCKTKoqgGInclANCndcMytnQRcX/Ap9cWzN+/XpOAUrXU+RJBU2NMqYPKi8jrDojHafYmpALQq2UDJ0dSDSTsLkgCQX1gwjeaBJSqxc6XCNwBXwqeLK7VVscm4u5m8HB38fbwJ/fDzL7WdNh4GPW+c+NRSjnc+RLBURH5Z5VF4kSpmTmkZeXSKdDP2aE4V1oSvNXTmvaur0lAKRdxvq+/LnElAAWD1N91WRsnR+JESXvh1bxms16+MO2gc+NRSlWZ8yWCK6osCif7dbvV2+j1YS2cHImT7P8d3u5tTfsHw/TDzo1HKVWlztfp3MmqDMRZsnJs/BGbSMtGLtitxNlkeLU92PKeHQwdCzfNcm5MSqkq5/KD18/KG5v4toEhzg2kquVkwcutC+YvfxoGP+a8eJRSTuPyieB/O63+hW7t38rJkVQhEXghrz+lJl3g3jXgAl2IKKVK5uJtJeHQyXQC6nnh4+UiOVEEni/0rIQmAaVcnksngsycXBJTswgPcaGniWf2K5iefkSTgFLKtRPB2axcAPqGuMhTs4sfg8Td1vTUGPDSzvWUUg5OBMaYYcaYXcaYWGPMtPNsN8YYI8aYcEfGc65VexLzz1+Vp3WOvSsg6kNr+o6fwd9Fm8oqpYpxWCIwxrgDM4HhQFdgvDGmawnb+QEPAn85KpbSLN5yBICrugRW9amr1sG/4IuR1vQN70DrS5wbj1KqWnHkFUE/IFZE9olIFjAPuLGE7f4FvAJkODCWEsWfOgtAqwCfqj51lWl+5FeYfbU149MYek90bkBKqWrHkU1lgoBDhebjgf6FNzDG9AJaishPxu4DG8wAACAASURBVJhSG7EbY6YAUwACAwOJjIysUECpqan2fbNtwvYj6QT6mAofr7oLPLacLrtnArC96+MkNB0EtbSshRX+O7sKLbNrcFSZHZkISrrxbh/jwBjjBrwB3F7WgURkFjALIDw8XCIiIioUUGRkJPn7bjhwEljD5CGdiBjSrkLHq9ZsufDPvAuwa1+jW9+7nBtPFSr8d3YVWmbX4KgyO/LWUDzQstB8MHCk0Lwf0B2INMbEAQOARVVVYbw3IQ2AnrVx/AER+KfVEuqMbztwoSSglLpwjkwEUUAHY0wbY4wXMA5YlL9SRE6LSGMRCRGREGAtcIOIrHdgTHbHT1tVEi0a1K2K01WdzNQiD4xt6vWiE4NRStUEDksEIpID3A/8AsQAC0RkuzHmn8aYGxx13vLKH5GsiV8dJ0dSieI3wItBBfNPJ2Fzr0XlU0o5hEP7VRCRJcCSc5Y9U8q2EY6M5Vypmbl4ebjVnh5Hf3sW/nzTmq5TH6Yd0KeGlVLl4iId7BS3ancCvVrVkvqB9JMFSWDMbOg+2rnxKKVqFJfsYiLXJmTl2mjVqJY8P/BK3shqnUZoElBKXTCXTATJ6VkANKrn5eRIKkHC7oLp8XOdF4dSqsZyyURwONl6orhjbRisfmZf6+eoD5wbh1KqxnLJRLAv7xmCGt+1xOb5BdOhY50Xh1KqRnPJRJBjsx5w9q1Tg+vK00/Cd1Os6b8t1BZCSqkKc8lEcOhkOgANfWpoHUHKkYIKYoD2VzgvFqVUjeeSieBUXmVxAx9PJ0dSAX++Ba93KZh/7rTzYlFK1QoumQi+33QYoOY9TLbhM/jtaWv60kc0CSilKoVLJoKm/t54uNWwe+oJu+DHB63pSx+BK59zZjRKqVqkBteWVlxOro1rQ5s7O4wLs+kL6+eQaTD0SefGopSqVVzyiiAj24Z7TboiEIHVb1vT/f/PubEopWodl0sE2bk2jqVk4OlWg4r+0ZUF0z6NnBeHUqpWqkGfhpXjQFINe5hszbtwOG+IhicOODcWpVSt5HKJ4I89iQC0b+rr5EjKYcsC+CWvPuCmD6FuLektVSlVrbhcIvjfzhMA9G9TzW+xiMDCu63pa16E0FucG49SqtZyuURwLG+IygbV/aniN0Otny0HwMB7nRuLUqpWc7lEcCT5LG0b13N2GOc3dzycPmhNj/3CubEopWo9l0sEnh5u1Xuc4m/vhl15o3vevRx8mzo3HqVUredyicBmE7o093d2GCXbtRS2LrCmJ/8CQX2cG49SyiW43JPFKRk51bN7iWXPwx+vW9OjPoBWA5wbj1LKZbhUIsgfh+D02WwnR1JITha80KRgfsR/IGyc8+JRSrkcl0oE2TbrZ7vq9AxB4STw8DZo0NJ5sSilXJJLJYKsXOunt0c1qRr5d4uC6WeTdZQxpZRTVJNPxKqRkmXdGsq/ReRUf7wB2VZ3FzywUZOAUsppXOqKIOmsdW+oef26zgvClguzIuDYFmt+8i8Q0M558SilXJ5LXRHkC2roxEQwf2JBEhjxH20dpJRyOpe6IkjNtm4JOa356I4fYNdia3r6EfCq5k84K6VcgktdESSdtRJBXS8njFVss8GCSdb0Nf9Pk4BSqtpwqUSQkddqKKiBE24NHVpr/XT3goH3Vf35lVKqFA5NBMaYYcaYXcaYWGPMtBLWTzXG7DDGbDHG/M8Y09qR8aRkWlcE3p5OuCL4YpT1c8zsqj+3Ukqdh8MSgTHGHZgJDAe6AuONMV3P2WwTEC4iocA3wCuOigcg21nNRle/AzlW99d0vs45MSilVCkceUXQD4gVkX0ikgXMA24svIGIrBCR9LzZtUCwA+PB3TjhttC2hfDrDGt6ykp9XkApVe04stVQEHCo0Hw80P88298J/FzSCmPMFGAKQGBgIJGRkRUKKCs7h5wsW4X3v1CeWacZtPoOAPa2vY1Du5Nhd9WcO19qamqVlbe60DK7Bi1z5XFkIijpq2+J92aMMX8DwoEhJa0XkVnALIDw8HCJiIioUEAzo5fiW8+biu5/QQp3Jtf5OtqNewtnPDYWGRlZNeWtRrTMrkHLXHkcmQjigcI9qAUDR87dyBhzJTADGCIimQ6Mh6xccK+qZwheDLJ+Nu0K476smnMqpVQFOLKOIAroYIxpY4zxAsYBiwpvYIzpBXwA3CAiJxwYCwDJmYKft6ejTwNZ6ZCbZU3/fbXjz6eUUhfBYYlARHKA+4FfgBhggYhsN8b80xhzQ95mrwK+wNfGmGhjzKJSDlcpTqTbaObv7chTQG42/L/m1nTfu7VyWClV7Tm0iwkRWQIsOWfZM4Wmr3Tk+c+VY4Pks1mOPcmXY6yfderDiFcdey6llKoELvVkcbYNujpyvOIjm2BfpDX9RJxeDSilagSXSQTpWTkApOWPTlOZRKwxh2dFWPODHwc3l/nVKqVqOJfpfTQ376nito0d0Nnb/L/Bzp+s6V4T4fIZlX8OpZRyEJdJBLa88YqNI27X5CeBJ+KgbsPKP75SSjmQy9y/sIl1ReBe2Xngmzutny37axJQStVILnNFkJuXCNwq84Gymf0hYac1feVzlXdcpZSqQi6TCLJzrXtDbpV1a+jM8YIk8Nge8G1aOcdVRWRnZxMfH09GRka596lfvz4xMTEOjKr60TK7hvKU2dvbm+DgYDw9y//wrMskgjMZVquhs5XVaui1jtbP0HGaBBwoPj4ePz8/QkJCyl2/c+bMGfz8/BwcWfWiZXYNZZVZREhKSiI+Pp42bdqU+7guU0eQd2eocgauj/uzYHrkexd/PFWqjIwMAgICHFPJr1QtY4whICDggq6gwYUSQX5l8UVXEeTmwKcjrOlbPtfnBaqAJgGlyq8i/y8u8ymWnwhK7h37ArzepWC6642lb6eUUjWEyySC/DxwUVcEm+ZAWl4nqc+cvOiYVPWXnJzMu+++W6F9R4wYQXJy8gXvFxYWxvjx44ssi4iIYP369fb5uLg4unfvbp9fv349gwcPplOnTnTu3Jm77rqL9PR0LtSGDRvo0aMH7du358EHH0Sk+BAip06dYtSoUYSGhtKvXz+2bdtmX7d06VI6depE+/bteemll+zLJ0yYQKdOnejevTuTJ08mOzsbgJ07dzJw4EDq1KnDf/7zn2Lnys3NpVevXlx3XcEQryLCjBkz6NWrF126dOGtt94qsk9UVBTu7u588803AKxYsYKePXvaX97e3nz//fcA3H777bRp08a+Ljo62n6OBx98kPbt2xMaGsrGjRvtx3/iiSfo3r073bt3Z/78+fbld955J2FhYYSGhjJmzBhSU1MBWLVqFb1798bDw8MeE0B0dDQDBw6kW7duhIaGFjlWfhk7duxYpIyLFy8mNDSUnj17Eh4ezh9//FHyH/JCiUiNevXp00cqYsuhZGn9xE/y2/ZjFdpflk4XedbfekXPq9gxnGDFihXODuGi7Nix44L3SUlJqbTz79+/X7p161biupycnEo7T74dO3ZI9+7dpUWLFpKammpfPmTIEImKiioxrmPHjknLli1l9erVIiJis9nk66+/lmPHLvy93rdvX1m9erXYbDYZNmyYLFmypNg2jz32mDz33HMiIhITEyOXX365iFi/j7Zt28revXslMzNTQkNDZfv27SIisnjxYrHZbGKz2WTcuHHy7rvviojI8ePHZd26dTJ9+nR59dVXi53rtddek/Hjx8u1115rXzZ79myZOHGiJCcn24+RLycnR4YOHSrDhw+Xr7/+utjxkpKSpGHDhpKWliYiIrfddluJ2y1evFiGDRsmNptN1qxZI/369RMRkZ9++kmuvPJKyc7OltTUVOnTp4+cPn1aRMT+U0TkkUcekRdffFFErL/V5s2bZeLEiUXOtWvXLtm9e7eIiBw+fFiaNWsmp06dKlLG3NzcImU8cuSI2Gw2ERHZvHmzdOrUqVjsIiX/3wDrpZTPVZdpNWSvI6jINdDiRyHqI2v6+rcgbGzlBabK7fkft7PjSEqZ2+Xm5uLu7l6uY3Zt4c+z13crdf20adPYu3cvPXv25KqrruLaa6/l+eefp3nz5kRHR7Njxw5GjhzJoUOHyMjI4KGHHmLKlCkAhISEsH79elJTUxk+fDiXXnopq1evJigoiB9++IG6dYs3XPjqq6+YOHEiMTExLFq0qNiVQUlmzpzJ+PHjGThwIGDdIx4zZky5yl/Y0aNHSUlJsR9n0qRJfP/99wwfPrzIdjt27ODJJ58EoHPnzsTFxXH8+HH27dtH+/btadu2LQDjxo3jhx9+oGvXrowYMcK+f79+/YiPjwegadOmNG3alMWLFxeLJz4+nsWLFzNjxgxef/11+/L33nuPr776Cre8f+amTQta7b399tuMHj2aqKioEsv4zTffMHz4cHx8fM77u/jhhx+YNGkSxhgGDBhAcnIyR48eZceOHQwZMgQPDw88PDwICwtj6dKl3HLLLfj7Wx1aighnz56136sPCQkBsMebr2PHjvbpFi1a0LRpUxISEmjQoEGpZfT19bUfNy0trdLqz1zm1lB+IjAXWkfwzeSCJPD3NdDntkqOTFVnL730Eu3atSM6OppXX7W6FV+3bh3//ve/2bFjBwCzZ89mw4YNrF+/nrfeeoukpKRix9mzZw/33Xcf27dvp0GDBnz77bclnm/+/PmMHTuW8ePHM3fu3HLFuG3bNnr16lXmdrt27Spyi6TwKzk5mcOHDxMcHGzfPjg4mMOHDxc7TlhYGAsXLrT/Lg4cOEB8fDyHDx+mZcuW590/OzubL774gmHDhpUZ78MPP8wrr7xS7AN07969zJ8/nyFDhjB8+HD27NkDwOHDh/nuu++45557Sj3mvHnziiXXGTNmEBoayiOPPEJmZqb9WCWVJSwsjJ9//pn09HQSExNZsWIFhw4VDM1+xx130KxZM3bu3MkDDzxQZhnzrVu3jqysLNq1a1ekjOHh4UXKCPDdd9/RuXNnrr32WmbPnl3uc5yPy1wRJJyx/sD5D5aVy/bvYFveP+zI9yGwqwMiU+V1vm/uhTm6fXm/fv2KtNF+6623+O677wA4dOgQe/bsISAgoMg++fehAfr06UNcXFyx40ZFRdGkSRNat25NcHAwkydP5tSpUzRs2LDEb34X+m2wU6dO9nvgJZES6gNKOse0adN46KGH6NmzJz169KBXr154eHiUa/97772XwYMHc9lll5031p9++ommTZvSp0+fYoO1Z2Zm4u3tzcqVK/ntt9+YPHkyv//+Ow8//DAvv/xyqVeDR48eZevWrVxzzTX2ZS+++CLNmjUjKyuLKVOm8PLLL/PMM8+UWparr76aqKgoLrnkEpo0acLAgQPx8Cj4GP3kk0/Izc3lgQceYP78+dxxxx3nLWd+XBMnTuSzzz6zJ738Mq5fv56FCxfaywgwatQoRo0axapVq3j66adZtmxZmecoi8tcEeQ/URzg61W+HfaugK9vt6ZvXwI9y75EV66hXr2CHmwjIyNZtmwZa9asYfPmzfTq1avENtx16tSxT7u7u5OTk1Nsm7lz57Jz505CQkJo164dKSkp9iuHgIAATp06Zd/25MmTNG7cGIBu3bqxadOmMuMu64ogODjYfssGrFszLVq0KHYcf39/PvnkE6Kjo/n8889JSEigTZs2BAcHF/l2fO7+zz//PAkJCUVu85Tmzz//ZNGiRYSEhDBu3DiWL1/O3/72N8D6dj569GjA+lDcsmULYFWYjxs3jpCQEL755hvuvfdee6UwwIIFCxg1alSRJ26bN2+OMYY6depwxx13sG7dOvs5SivLjBkziI6O5rfffkNE6NChQ5HY3d3dGTt2bKlXfYWlpKRw7bXX8sILLzBgwAD78tLKWNjgwYPZu3cviYmJZZ6nLC6TCE6mWSOT+XiV4yIo6mP4YqQ13fduCBnkwMhUdebn58eZM2dKXX/69GkaNmyIj48PO3fuZO3atRU6j81m4+uvv2bLli3ExcURFxfHDz/8YL89FBERwZw5c+zfVD/77DOGDh0KwP3338/cuXP566+/7MebM2cOx44dK3KO/CuCkl4NGjSgefPm+Pn5sXbtWkSEzz//nBtvLN5EOjk5maws6//po48+YvDgwfj7+9O3b1/27NnD/v37ycrKYt68edxwww327X755Rfmzp1b7FZPSV588UXi4+OJi4tj3rx5XH755cyZMweAkSNHsnz5cgBWrlxpv9e+f/9+++9uzJgxvPvuu4wcOdJ+zLlz5xa7LXT06FHAuhr6/vvv7S2xbrjhBj7//HNEhLVr11K/fn2aN29Obm6u/dbfli1b2LJlC1dffTUiQmxsrP1YP/74I507dz5vGbOyshg1ahSTJk3i5ptvLrKutDLu3bvX/h7YuHEjWVlZxa4+K8Jlbg2dzba6lqjjUcabMPkgLJ5qTUc8CRHTHByZqs4CAgIYNGgQ3bt3Z/jw4Vx77bVF1g8bNoz333+f0NBQOnXqVORb3YVYtWoVQUFBBAUF2ZcNHjyYHTt2cPToUaZMmcLOnTsJCwvDGEN4eDgvvvgiAIGBgcyePZvHHnuMEydO4ObmxuDBg7npppsuOI733nuP22+/nbNnzzJ8+HB7RfH7778PwD333ENMTAyTJk3C3d2drl278vHHHwPg4eHBO++8wzXXXENubi6TJ0+mW7du9v1at25tr4i+6aabeOaZZzh27Bjh4eGkpKTg5ubGm2++yY4dO+wVryWZNm0aEyZM4LXXXsPf35+PPvqozHLFxcVx6NAhhgwZUmT5hAkTSEhIQETo2bOnvZwjRoxgyZIltG/fHh8fHz755BPAquPIv63l7+/PnDlz8PDwwGazcdttt5GSkoKIEBYWxnvvWb0OREVFMWrUKE6dOsWPP/7Is88+y/bt21mwYAGrVq0iKSmJTz/9FIBPP/2Unj172sv4xhtv4Ovray/jokWLmDBhAp6entStW5f58+dXSoWxKeleWHUWHh4uhdtTl9ectQd46vttrJt+BU1LG8BeBJ5vYE036wH3VFIbXSeKjIwkIiLC2WFUWExMDF26dCl7w0K0DxrXoGUuXUn/N8aYDSISXtL2LnNryJ7uzpc8lz5ZMF0LkoBSSpWHyyQCymo+uvRJ+CuvA7mHt1ZRUEop5XwuU0dg72mopDzwzeSCZqLXvgYNWlVVWEop5XSukwjyMkGxPBDzY0ESmBoD/sWbyymlVG3mMreG8ivFi9Swx/0B8622yVz9giYBpZRLcp1EkPezyBXBp3lNATsOh0vK/zi4UkrVJq6TCPJvDeVngj/etH561IVb5zklJlX9JSUl2Z++bdasGUFBQfb5/IeqymP27NnFHvAqLCsri0aNGvH0008XWR4cHFykK+tly5YVeUhq8eLF9OnTh759+9K5c2eeeOKJCyhdgSVLlti7j87vU6kkc+fOpWvXrnTr1o1JkyYBkJOTg7u7u/33MmrUKPv2IsK0adPs3SnPnDkTgIULF9q7U+7bty+rV68GrIfCevfuTc+ePenevTsffvih/Vhz5syhR48ehIaGMmLECE6etLqCnzp1Kp06dSI0NJTRo0dz+vRp+z7R0dEMGDCAbt260aNHD3v315mZmdx11132brsLP4EMVp9Exhh7lxyxsbHUrVvXXsb77ruvzLjyvfTSSxhj7H/H7du327vefvPNN4ts+/rrr9OtWze6devG22+/bV+eX8aBAwcWK2OlKK1b0ur6qmg31B/9vk9aP/GTJKdliZxNLuhSOjG2QserKbQb6srz7LPPlthVcnkMGjRINm3aVOr6H374QQYNGiQdOnQosjwoKMjeNbGIyG+//SY33nijiIhER0dL+/btZdeuXZKSkiLZ2dn27p0vRFZWlrRp00bi4uIkIyNDunfvLrt27Sq2XUxMjPTu3dseT37XyNnZ2VK/fv0Sjz1r1iy544477F0n5+9z5swZ+7INGzbYu9TOyMiQjIwMEbG6dW7ZsqUcP35cMjMzpUmTJpKUlCQiVjfPTz31lIiILF26VLKzs0VEZOrUqTJ9+nR7ubp37y5btmwREZGEhAR7t87Tp0+XZ599VkREcnNzJTEx0R7z6dOn5bLLLpPw8HD732zPnj0SFhZWrHwlxfWvf/3Lvn7//v0ybNiwIn/HY8eOSVRUlDzxxBPyxhtv2LfdtGmThIaGSnp6umRlZUlERITs27evSBlTUlKKlLE02g11KaRwbfFLea2CGoZAQDtnhaQu1M/T4FjZTXvr5uaAeznf2s16wPCXyt6uBJ999hkzZ84kKyuLSy65hHfeeQebzcYdd9xBdHQ0IsKUKVMIDAwkOjqasWPHUrduXdatW4eXV9E+r+bOncvUqVN54403iIqKom/fvmWe/+WXX+bpp5+mY8eOnDlzBg8PD/7+979fcDnWrl1Lly5daN26NQC33HILP/zwA//4xz+KbDdr1iweeOABGjSwHros3P1zad577z0WLlxor5sr3J1yvsLdKRfukykzMxObzVbkAystLY2GDRty5swZ2rdvD1CkE7kBAwbw008/AfDzzz/Tp08fevToAWDvmwmsJ3jzu4Rwc3Mr0k3D9OnTmT59Oi+88EKZ5SsprsIDBj3yyCO8+uqrRXpbDQwMJDAwsNhVSExMDAMHDrR3Tz548GC+++47pk6dWmoZK4vL3BrK57PoroKZB8ruqEupkmzbto3vvvuO1atXEx0dTU5ODvPmzWPDhg0kJiaydetWtm3bxqRJkxg7diw9e/Zk/vz5REdHF0sCaWlprFy5khEjRlxw99N9+vQpc7tly5aV2NFcflcJ5ek+GmD37t3ExMQwaNAgBg4cyK+//lqkDH369GHgwIH8+OOP9uX79+9nzpw5hIeHM2LECPbu3Wtf980339CpUydGjhxZpJuIuLg4QkNDad26NU899RSBgYHUqVOHd955h65du9KiRQtiY2OZMGFCsRhnz55t7xZj9+7diAhXX301vXv35rXXXgMgMTERLy8vnnzySXr37s3YsWNJSEgArO4gTpw4UWI32bGxsfTq1YuIiAj7rayS4rr99tsB+Pbbb2nbtm2RxHA+PXr0IDIykpMnT5KWlsbPP/9cpOO7kspYWVzoigA8ycEzJi8LPxarA8/XNOX85n62CroeWLZsGVFRUYSHW0/snz17lpYtW3LNNdewa9cuHnroIUaMGMHVV19d5rEWLVrEVVddhbe3NzfffDPh4eH85z//wc3NrVK6n77yyisrpfvpnJwc9u3bx8qVKzlw4ABDhgxhx44d+Pn5ceDAAfsH4RVXXEGPHj0ICQkhIyMDPz8/1q9fz4IFC7jrrrtYsWIFAGPGjGHMmDGsWLGCp59+2p5YQkJC2LJlC4cPH2bkyJGMGTMGf39/PvjgA7Zs2ULr1q259957efPNN3n22Wft8T3//PP4+voybtw4e7x//vknf/31F97e3gwdOpTw8HA6depEXFwcQ4cO5c033+SVV17h8ccf5+OPP+bRRx+1d25XWHBwMAcPHqRRo0asW7eO0aNHExMTg5eXV7G4XnnlFe6//35eeeWVC+oiunv37kydOpUrr7wSX19fe/fehb344otFylhZHPpJaIwZZozZZYyJNcYU673NGFPHGDM/b/1fxpgQR8UiCMu8HitY4NvEUadSLkBEmDx5sr33zl27dvH0008TEBDAli1buPTSS3nrrbf4v//7vzKPNXfuXJYuXUpISAh9+/blxIkTrFq1Cii7++kNGzaUefyyrgjK6j46X3BwMDfeeCMeHh60a9eOdu3asXfvXowx9u3bt2/PZZddZk88QUFB9u6UR48eXWJCGjp0KDExMcXGdw4KCqJz58788ccfbNiwAU9PT9q0aYObmxu33HJLkd5WP/74Y3799Ve++OKLIvFGREQQEBBAvXr1GD58OBs3bqRp06b4+PjYe0a9+eab2bhxI8nJyezYsYPBgwfbR5cbMWIEmzZtwtvbm0aNGgHWeBStW7cmNja2xLhWr15NbGws+/fvtyfEY8eOERoaar/yKM2UKVPYuHEjq1aton79+kW6uP74449Zvnx5kTJWmtIqDy72BbgDe4G2gBewGeh6zjb3Au/nTY8D5pd13IpWFr+/Yk9BBXF2ZoWOURNpZXHlKVxZvGXLFunYsaMkJCSIiEhiYqIcOHBATpw4YT9/VFSU5L9fhw0bJqtWrSp2zJMnT0rTpk0lKyvLvmzWrFkyZcoUERF56KGH5PnnnxcRq1L2hhtukC+//FJERDZu3CgdOnSQPXv2SEpKiuTk5Mhrr712weXKysqSkJCQIpXFO3fuLLbdjz/+KJMnTxYRq9I3vwI0KSnJXsF74sQJadu2rX3/Rx99VD777DMRsSq6BwwYICJW5Wt+ZXFUVJQEBweLiMihQ4fk7Nmz9t9pu3btZPv27XLw4EFp3ry5vVL3ySeflIcfflhErHGEu3btWqTCN3//3r17S3p6umRnZ0tERIQsXbpURERGjx4tK1euFBGRDz/8UMaNG1esvIUr+E+cOGEfo3rPnj3SvHlzSU5OLjGuxx9/vNixzq30FxGZMWNGkcri/N+riFXJ3LFjR/u4zPll3L9/f7Fjl+RCK4sdmQgGAr8Umn8SePKcbX4BBuZNewCJ5PWIWtqroolg2bw3RZ71l6zF0yq0f02liaDynNtq6Msvv5SwsDDp0aOH9O7dW9atWycbNmyQnj17SlhYmPTs2VN++eUXERGZP3++dOzYUcLCwiQzs+CLyIcffigTJkwocp4TJ05IkyZNJCsrS06dOiVjx46V0NBQ6dGjhzzxxBP2D1ARq7VRr169pGPHjtKlSxeZNq1i7+9FixZJhw4dpG3btvLSSy/Zl0+fPl0WL14sIiI2m00efPBB6dKli/To0UMWLFggIiKrVq2S7t2722P85JNP7PsnJSXJsGHDpHv37jJw4EB7C55///vf0rVrVwkLyv8kkwAAB5BJREFUC5OBAwfKn3/+KSIiP//8c5FjffTRR/ZjzZw5Uzp37iw9evSQ66+/Xg4cOCAiIq1bt5aWLVtKWFiYhIWFyb333mvf59NPP5UuXbpIt27divxu9u3bJ5deeqn06NFDrrjiCjl48GCx30nhRDB//nx7vL1795affvqp1LhOnjxZ7FiFE8GhQ4ckKChI/Pz8pH79+hIUFCRpaWkiIjJw4EDp0qWLhIWFFfnfzS9jjx49ipWxJBeaCBzWDbUxZgwwTETuypufCPQXkfsLbbMtb5v4vPm9edsknnOsKcAUgMDAwD7z5l14u//TsavpdmQBh/s9i3g3rGixapzU1NQiLTRqmvr169tbh5TXhQxeX1tomV1DecscGxtb7FmDoUOHltoNtSMri0uq0To365RnG0RkFjALrPEIKtS/fkQEkZGX1Oi++SuiNoxHcKEVv9pPvWvQMpfO29ubXr16lfu4jqwsjgdaFpoPBo6Uto0xxgOoD5xEKaVUlXFkIogCOhhj2hhjvLAqgxeds80i4La86THAcnHUvSpVY+lbQqnyq8j/i8MSgYjkAPdjVQjHAAtEZLsx5p/GmBvyNvsYCDDGxAJTAR0gWBXh7e1NUlKSJgOlykFESEpKwtu7lOF4S+HQB8pEZAmw5JxlzxSazgBudmQMqmYLDg4mPj6+zPbXhWVkZFzwP0JNp2V2DeUps7e3N8HBwRd0XJd5sljVTPkP61yIyMjIC6ooqw20zK7BUWXWPhaUUsrFaSJQSikXp4lAKaVcnMOeLHYUY0wCcKCCuzfG6sbClWiZXYOW2TVcTJlbi0iJvW3WuERwMYwx60t7xLq20jK7Bi2za3BUmfXWkFJKuThNBEop5eJcLRHMcnYATqBldg1aZtfgkDK7VB2BUkqp4lztikAppdQ5NBEopZSLq5WJwBgzzBizyxgTa4wp1qOpMaaOMWZ+3vq/jDEhVR9l5SpHmacaY3YYY7YYY/5njGntjDgrU1llLrTdGGOMGGNqfFPD8pTZGHNL3t96uzHmq6qOsbKV473dyhizwhizKe/9PcIZcVYWY8xsY8yJvBEcS1pvjDFv5f0+thhjel/0SUsbw7KmvgB3YC/QFvACNgNdz9nmXuD9vOlxwP9v715C46riOI5/f1LFqlVLgyBaSJFGlC5UitSFNVIRKSXduFAoWikuCroQERcuKrpQfOBKqIriY+ETqkGQblQSixErCWJ9lFBLDSiCaBa2vtq/i3Mrw+QxJ0zuvc7c3wcC9565ZP7/zCT/OefcnPNm3XFXkPONwDnF8e4m5FxctwoYAyaAjXXHXcHrvB6YBFYX5xfVHXcFOT8P7C6OrwSO1h13lzlvBq4Bvlrg8a3AB6QdHjcBn3X7nP3YI7gWmI6IIxHxF/AGsL3tmu3AK8XxO8AWSfNtm9krOuYcER9FxPHidIK0Y1wvy3mdAR4FngD+qDK4kuTkfDfwbET8ChARP1cc43LLyTmA84vjC5i7E2JPiYgxFt+pcTvwaiQTwIWSLu7mOfuxEFwC/NByPlO0zXtNpA10ZoE1lURXjpycW+0ifaLoZR1zlnQ1sDYi3q8ysBLlvM5DwJCkA5ImJN1SWXTlyMn5YWCHpBnS/if3VhNabZb6+95RP+5HMN8n+/Z7ZHOu6SXZ+UjaAWwEbig1ovItmrOkM4BngJ1VBVSBnNd5BWl4aJjU6xuXtCEifis5trLk5Hw78HJEPC3pOuC1IudT5YdXi2X/+9WPPYIZYG3L+aXM7Sr+d42kFaTu5GJdsf+7nJyRdBPwEDASEX9WFFtZOuW8CtgAfCzpKGksdbTHJ4xz39vvRcTfEfE98B2pMPSqnJx3AW8BRMSnwNmkxdn6Vdbv+1L0YyH4HFgvaZ2ks0iTwaNt14wCdxbHtwIfRjEL06M65lwMkzxHKgK9Pm4MHXKOiNmIGIiIwYgYJM2LjETEwXrCXRY57+13STcGIGmANFR0pNIol1dOzseALQCSriAVgvy9TXvPKHBHcffQJmA2In7s5hv23dBQRPwj6R5gP+mOg5ci4pCkR4CDETEKvEjqPk6TegK31Rdx9zJzfhI4D3i7mBc/FhEjtQXdpcyc+0pmzvuBmyV9DZwEHoiIX+qLujuZOd8PvCDpPtIQyc5e/mAn6XXS0N5AMe+xBzgTICL2kuZBtgLTwHHgrq6fs4d/XmZmtgz6cWjIzMyWwIXAzKzhXAjMzBrOhcDMrOFcCMzMGs6FwCyTpJOSplq+BiUNS5otVr78RtKe4trW9m8lPVV3/GYL6bv/IzAr0YmIuKq1oVjCfDwitkk6F5iSdHpto9PtK4FJSfsi4kC1IZt15h6B2TKJiN+BL4DL2tpPAFN0uTCYWVlcCMzyrWwZFtrX/qCkNaQ1jQ61ta8mrfczVk2YZkvjoSGzfHOGhgrXS5oETgGPF0sgDBftXwKXF+0/VRirWTYXArPujUfEtoXaJQ0BnxRzBFNVB2fWiYeGzEoWEYeBx4AH647FbD4uBGbV2AtslrSu7kDM2nn1UTOzhnOPwMys4VwIzMwazoXAzKzhXAjMzBrOhcDMrOFcCMzMGs6FwMys4f4FlXHCdKZoR3wAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Best alpha is found from above analysis of TfDIF. using best Alpha, test accuracy is calcuated as around.\n",
    "Best_alpha=0.01\n",
    "NB_best_validation(X_tr_tfidf,y_train,X_cr_tfidf,y_cv,Best_alpha,features_tfidf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "NeWORPrbU8Fi"
   },
   "source": [
    "### 3.4 <font color='red'> Alpha-Best   <font color='blue'>  on - categorical, numerical features + project_title + preprocessed_essay ,<font color='red'> SET 4</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 436
    },
    "colab_type": "code",
    "id": "Nilnvn5nU8Fi",
    "outputId": "300e3d7f-8672-43cd-e03e-04e7104ba0fc"
   },
   "outputs": [],
   "source": [
    "# Best alpha is found from above analysis of Std. using best Alpha, test accuracy is calcuated as around.\n",
    "Best_alpha= 0.066\n",
    "NB_best_validation(X_tr,y_train,X_cr,y_cv,Best_alpha,features_std)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "CQM3zu-iU8Fk"
   },
   "source": [
    "<h1>3. Conclusions</h1>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "miDxCezsU8Fl"
   },
   "source": [
    "## Summary of above program as below:\n",
    "### Step 1: Import the necessary Libraries\n",
    "we will need to import libraries that allow for data analysis and data visualization to get acclimated to the dataset. We will be using pandas, numpy, matplotlib and seaborn to conduct this. Data Exploration libraries\n",
    "\n",
    "### Step 2: Read in the dataset.\n",
    "We will use the pandas .read_csv() method to read in the dataset. Then we will use the. head() method to observe the first few rows of the data, to understand the information better. In our case, the feature(column) headers tell us pretty little. This is fine because we are merely trying to gain insight via classifying new data points by referencing it’s neighboring elements.\n",
    "\n",
    "### Step 3: Standardize (normalize) the data scale to prep for Multinomial NB algorithm.\n",
    "Because the distance between pairs of points plays a critical part on the classification, it is necessary to normalize the data This will generate an array of values. Again,  Naive Bayes depends on the distance between each feature. Please see Section 1 for all normalization.\n",
    "\n",
    "### Step 4: Split the normalized data into training and test sets.\n",
    "This step is required to prepare us for the fitting (i.e. training) the model later. The “X” variable is a collection of all the features. The “y” variable is the target label which specifies the classification of 1 or 0 based. Our goal will be to identify which category the new data point should fall into.\n",
    "\n",
    "Please see functions as covered below, used in above program: def multinomialNB_validation_alpha_analysis: def NB_best_validation\n",
    "\n",
    "### Step 5: Create and Train the Model.\n",
    "Here we create a Naive Bayes Object and use the .fit() method to train the model. Upon completion of the model we should receive confirmation that the training has been complete\n",
    "\n",
    "Please see functions as covered below, used in above program: def multinomialNB_validation_alpha_analysis def NB_best_validation\n",
    "\n",
    "### Step 6: Make Predictions.\n",
    "Here we review where our model was accurate and where it misclassified elements.\n",
    "\n",
    "Please see functions as covered below, used in above program: def multinomialNB_validation_alpha_analysis def NB_best_validation\n",
    "\n",
    "### Step 7: Evaluate the predictions.\n",
    "Evaluate the Model by reviewing the classification report or confusion matrix. By reviewing these tables, we are able to evaluate how accurate our model is with new values.\n",
    "\n",
    "def multinomialNB_validation_alpha_analysis def NB_best_validation:\n",
    "\n",
    "### Setp 8:Classification Report :\n",
    "This tells us our model was around 84% accurate… Print out classification report and confusion matrix\n",
    "\n",
    "print(classification_report(y_test, pred))\n",
    "\n",
    "I have covered various set to show confusion matrix.\n",
    "\n",
    "Please see section 2. covered various data sets and created confusion matrix.\n",
    "\n",
    "### Step 9: Evaluate alternative Alpha for better predictions.\n",
    "To simplify the process of evaluating multiple cases of Alpha values, we create a function to derive the error using the average where our predictions were not equal to the test values.\n",
    "\n",
    "Please see section 2. covered various data sets and created error accuracy reports.\n",
    "\n",
    "### Step 10: Plot Error Rate\n",
    "Here we see that the error rate continues to decrease as we increase the Alpha. A picture tells a thousand words. Or at least here, we are able to understand what value of Alpha leads to an optimal model. The Alpha of 1 or 0.9 seems to give a decent error rate without too much noise.\n",
    "\n",
    "### Step 11: Adjust Alpha value per error rate evaluations \n",
    "This is just fine tuning our model to increase accuracy. We will need to retrain our model with the new Alpha.\n",
    "Please see section 3 in above program. we have created confusion matrix for optimal Alpha value for various data sets. As we can see for optimal Alpha, Accuracy is much higher - so prediction is much better.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "raw",
    "id": "xmCuubPvU8Fl"
   },
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [
    "BKMTtTRsU8Ca",
    "T9fPeXrCU8Cg",
    "3fKZ7Vj2U8Cw",
    "rEOuX_qsU8C9",
    "wJj8hTVpU8DY",
    "_MRmjcyoU8DZ",
    "JAGpFxofU8Db",
    "Sivr5hi9U8De",
    "8zI0Iq8iU8Dn",
    "67A3UkycU8Dr",
    "B5hVdV2EU8D2",
    "2UMRsjsEU8EB",
    "_kIodqX7U8EH",
    "vDucjqqgU8EJ",
    "rQhPrFajU8EL",
    "PBUXhlNxU8ER",
    "N5-0KZL-U8EV",
    "Nl7sCl4OU8Ea",
    "cRZMkojYU8E5",
    "6DxjXFKoU8FC",
    "OGHD6xEVU8FD",
    "Ej0C09rcU8FI",
    "ioZ8czwmU8FL",
    "L-X5RIXTU8FR",
    "XMZhiZcQU8FV",
    "ecObAWXmU8FW",
    "QCGuXvfLU8Fb",
    "bpfIpOEIU8Fe",
    "NeWORPrbU8Fi",
    "miDxCezsU8Fl"
   ],
   "name": "Full_Naive_Bayes_Analysis.ipynb",
   "provenance": [],
   "toc_visible": true,
   "version": "0.3.2"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}