Spam-Filter detetction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Exercise:\n",
    "\n",
    "* There are 922 e-mails divided equally between spam and non-spam(ham)\n",
    "* The first line of every email contains the subject and the third line contains the body as shown:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![title](sample_email.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Perform text analysis only on the body to detect spam mails"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Steps:\n",
    "1. Read the body of all the emails. (Hint: Just google how to read multiple text files in python. Its very simple - all the text files can be read at once using a simple loop. But make sure while reading itself you read only the body (third line) and not the subject).\n",
    "2. Once you read, the content of all the mails can be collected in a python list.\n",
    "3. Create a pandas dataframe from that list and add the class label column [spam(1) or ham(0)]\n",
    "4. The shape of the pandas dataframe must be (922 x 2) - 922 rows and 2 columns\n",
    "5. Now perform the standard preprocessing steps on the contents of the mail (stopword removal, lower casing, etc.)\n",
    "6. After that create vectors using Binary BOW, count BOW and TFIDF.\n",
    "7. Use BernoulliNB with Binary BOW and MultinomialNB with Count BOW and TFIDF\n",
    "8. Evaluate all the three methods using Accuracy, Confusion Matrix and F-1 Score.\n",
    "9. Compare the results of the three methods"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "\n",
    "import sqlite3\n",
    "import pickle\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.naive_bayes import BernoulliNB\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "\n",
    "\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from matplotlib.colors import ListedColormap\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "\n",
    "import re\n",
    "from nltk.corpus import stopwords\n",
    "import nltk \n",
    "\n",
    "import glob\n",
    "import errno\n",
    "\n",
    "from sklearn.manifold import  TSNE\n",
    "from sklearn.utils import shuffle"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### reading multiple .txt file from Spam and Ham folder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def ret(loc):\n",
    "    path =loc\n",
    "    files = glob.glob(path)\n",
    "    lst=[]\n",
    "    for fname in files:\n",
    "        try:\n",
    "            with open(fname,'r') as f:\n",
    "                lines = f.readlines()\n",
    "                str = '' \n",
    "                for i in range(2,len(lines)):\n",
    "                    str += lines[i].rstrip('\\n') \n",
    "                    lst.append(str)\n",
    "                f.close()\n",
    "        except IOError as exc:\n",
    "            if exc.errno != errno.EISDIR:\n",
    "                raise\n",
    "    return lst"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "481\n",
      "481\n"
     ]
    }
   ],
   "source": [
    "spam=ret('/Users/abc/Desktop/machine learning/Models vectors/spam filter/Spam Dataset/Spam emails/*.txt')\n",
    "ham=ret('/Users/abc/Desktop/machine learning/Models vectors/spam filter/Spam Dataset/Ham emails/*.txt')\n",
    "print(len(spam))\n",
    "print(len(ham))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Making pandas Dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mail</th>\n",
       "      <th>type</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>* * * * * * * * * * * * * * * display box cred...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>dear nlpeople , ' m sure ' ll agree auto insur...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>want best hunt camp vacation life , felton 's ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>57 million email addresses $ 99 want money ? p...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>attention ! warn ! adult ! warn ! adult ! unde...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>attention ! warn ! adult ! warn ! adult ! unde...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>subject : re : are debt ? help . qualify finge...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>= = = = = = = = = = = = = = = = = = = = = = = ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>57 million email addresses $ 99 want money ? p...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>$ 3500 per week home computer ! put free softw...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>2 newest hottest interactive adult web site so...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>hi . name denni . third-generation , professio...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>desk robert g . allen dear nlpeople first let ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>* ever speed ticket need desperate ? * become ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>special : work home , sell successfully earn t...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>rumor are true ! * * * * * * * * * * * * * * *...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>federally mandate , electric deregulation , al...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>pardon intrusion . offence mean . interest , s...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>was released ! ! introducing . . . millions vo...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>ever treat bid embarrase credit report bid cre...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>: pay off 30 mortgage much less pay off credit...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>authenticate sender subject : ( + ( + bull _ s...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>hey , ' ll point . web site free trial 10 best...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>horizon publish invite saltwater aquarium hobb...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>multi-part message mime format . - - part0 _ 9...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>read 1st paragraph ! convince , delete . bet i...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>released . . . 30 , 0 , 0 email address ! plus...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>attention ! warn ! adult ! warn ! adult ! unde...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>hello , name kevin elphick manage director nur...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>b r g n r f r e s 1 - stop travel supplier air...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>932</th>\n",
       "      <td>* deadline abstracts - april 30 , 1998 * tag t...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>933</th>\n",
       "      <td>certamen phonologicum iii pier marco bertinett...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>934</th>\n",
       "      <td>second call papers icslp ' 98 5th internationa...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>935</th>\n",
       "      <td>pre - final programme call participation works...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>936</th>\n",
       "      <td>call participation logic course internet : dyn...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>937</th>\n",
       "      <td>follow list contents forthcome issue journal l...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>938</th>\n",
       "      <td>association language awareness knowledge langu...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>939</th>\n",
       "      <td>evaluation parsing systems workshop jointly or...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>940</th>\n",
       "      <td>= = = = = = = = = = = = = = = = = = = = = = = ...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>941</th>\n",
       "      <td>tenth conference nordic general linguistic hol...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>942</th>\n",
       "      <td>change date extended deadline - - - - - - - - ...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>943</th>\n",
       "      <td>acl / coling-98 ( association computational li...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>944</th>\n",
       "      <td>december 17 18 , 1998 , tilburg university hos...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>945</th>\n",
       "      <td>final call papers als-98 australian linguistic...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>946</th>\n",
       "      <td>call papers . student conference linguistic 10...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>947</th>\n",
       "      <td>special \" stream \" 20th annual meet cognitive ...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>948</th>\n",
       "      <td>workshop mathematical linguistics ( wml ) penn...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>949</th>\n",
       "      <td>1998 nic symposium intercultural communication...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>950</th>\n",
       "      <td>invitation workshop during 31st sle conference...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>951</th>\n",
       "      <td>book list below linguist office available revi...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>952</th>\n",
       "      <td>chus - - phonetic phonology group university c...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>953</th>\n",
       "      <td>call papers distribute access linguistic resou...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>954</th>\n",
       "      <td>coling-acl ' 98 workshop call papers + - - - -...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>955</th>\n",
       "      <td>call paper syllable : typology theory conferen...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>956</th>\n",
       "      <td>announcement call papers january 1998 - - - - ...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>957</th>\n",
       "      <td>call papers twenty after publication linguisti...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>958</th>\n",
       "      <td>call participation towards european evaluation...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>959</th>\n",
       "      <td>final call paper lfg98 1998 international lexi...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>960</th>\n",
       "      <td>call papers 5th international conference spoke...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>961</th>\n",
       "      <td>iii brazilian international conference cogniti...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>962 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  mail  type\n",
       "0    * * * * * * * * * * * * * * * display box cred...     1\n",
       "1    dear nlpeople , ' m sure ' ll agree auto insur...     1\n",
       "2    want best hunt camp vacation life , felton 's ...     1\n",
       "3    57 million email addresses $ 99 want money ? p...     1\n",
       "4    attention ! warn ! adult ! warn ! adult ! unde...     1\n",
       "5    attention ! warn ! adult ! warn ! adult ! unde...     1\n",
       "6    subject : re : are debt ? help . qualify finge...     1\n",
       "7    = = = = = = = = = = = = = = = = = = = = = = = ...     1\n",
       "8    57 million email addresses $ 99 want money ? p...     1\n",
       "9    $ 3500 per week home computer ! put free softw...     1\n",
       "10   2 newest hottest interactive adult web site so...     1\n",
       "11   hi . name denni . third-generation , professio...     1\n",
       "12   desk robert g . allen dear nlpeople first let ...     1\n",
       "13   * ever speed ticket need desperate ? * become ...     1\n",
       "14   special : work home , sell successfully earn t...     1\n",
       "15   rumor are true ! * * * * * * * * * * * * * * *...     1\n",
       "16   federally mandate , electric deregulation , al...     1\n",
       "17   pardon intrusion . offence mean . interest , s...     1\n",
       "18   was released ! ! introducing . . . millions vo...     1\n",
       "19   ever treat bid embarrase credit report bid cre...     1\n",
       "20   : pay off 30 mortgage much less pay off credit...     1\n",
       "21   authenticate sender subject : ( + ( + bull _ s...     1\n",
       "22   hey , ' ll point . web site free trial 10 best...     1\n",
       "23   horizon publish invite saltwater aquarium hobb...     1\n",
       "24   multi-part message mime format . - - part0 _ 9...     1\n",
       "25   read 1st paragraph ! convince , delete . bet i...     1\n",
       "26   released . . . 30 , 0 , 0 email address ! plus...     1\n",
       "27   attention ! warn ! adult ! warn ! adult ! unde...     1\n",
       "28   hello , name kevin elphick manage director nur...     1\n",
       "29   b r g n r f r e s 1 - stop travel supplier air...     1\n",
       "..                                                 ...   ...\n",
       "932  * deadline abstracts - april 30 , 1998 * tag t...     0\n",
       "933  certamen phonologicum iii pier marco bertinett...     0\n",
       "934  second call papers icslp ' 98 5th internationa...     0\n",
       "935  pre - final programme call participation works...     0\n",
       "936  call participation logic course internet : dyn...     0\n",
       "937  follow list contents forthcome issue journal l...     0\n",
       "938  association language awareness knowledge langu...     0\n",
       "939  evaluation parsing systems workshop jointly or...     0\n",
       "940  = = = = = = = = = = = = = = = = = = = = = = = ...     0\n",
       "941  tenth conference nordic general linguistic hol...     0\n",
       "942  change date extended deadline - - - - - - - - ...     0\n",
       "943  acl / coling-98 ( association computational li...     0\n",
       "944  december 17 18 , 1998 , tilburg university hos...     0\n",
       "945  final call papers als-98 australian linguistic...     0\n",
       "946  call papers . student conference linguistic 10...     0\n",
       "947  special \" stream \" 20th annual meet cognitive ...     0\n",
       "948  workshop mathematical linguistics ( wml ) penn...     0\n",
       "949  1998 nic symposium intercultural communication...     0\n",
       "950  invitation workshop during 31st sle conference...     0\n",
       "951  book list below linguist office available revi...     0\n",
       "952  chus - - phonetic phonology group university c...     0\n",
       "953  call papers distribute access linguistic resou...     0\n",
       "954  coling-acl ' 98 workshop call papers + - - - -...     0\n",
       "955  call paper syllable : typology theory conferen...     0\n",
       "956  announcement call papers january 1998 - - - - ...     0\n",
       "957  call papers twenty after publication linguisti...     0\n",
       "958  call participation towards european evaluation...     0\n",
       "959  final call paper lfg98 1998 international lexi...     0\n",
       "960  call papers 5th international conference spoke...     0\n",
       "961  iii brazilian international conference cogniti...     0\n",
       "\n",
       "[962 rows x 2 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "spam=pd.DataFrame(spam,columns=['mail'])\n",
    "ham=pd.DataFrame(ham,columns=['mail'])\n",
    "spam['type']=1\n",
    "ham['type']=0\n",
    "data=pd.concat([spam,ham],axis=0,ignore_index=True)\n",
    "data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Loading Data file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<function Connection.close>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conn=sqlite3.connect('spam.sqlite')\n",
    "c=conn.cursor()\n",
    "spam.to_sql('Data',conn,if_exists='replace',index=False)\n",
    "conn.close"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Checking Duplicated mails"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(937, 2)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=data.drop_duplicates(subset=['mail'])\n",
    "data.shape #25 duplicates mail "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Text Preprocessing "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Removal of stopwords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stop = set(stopwords.words('english'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'am', 'itself', 'will', 'on', 'himself', \"wasn't\", 'that', 'after', 'being', 'do', 'their', 'but', 'don', 'when', 'll', 'wouldn', 'under', 'his', 'out', 'whom', 'as', \"mightn't\", 'ain', 'him', 'is', 'off', 'themselves', 'there', 'then', 're', \"that'll\", 'for', 'my', 'an', 'down', 'our', 'up', 'they', 'who', 'have', 'such', 'in', 'same', 'its', 'of', \"you'll\", 'aren', 'them', 'above', 'did', 'be', 'again', \"shouldn't\", 'into', \"it's\", 'he', 'hers', 'yourself', \"haven't\", 'other', 'below', 'so', \"you're\", \"weren't\", 'been', 'needn', 'all', \"shan't\", \"aren't\", 'by', 'no', 'ourselves', 'mustn', 'each', 'until', 'shan', \"mustn't\", \"you've\", 'were', 'which', 'very', 'd', 'about', 'are', 'hadn', 'she', 'a', 'both', 'where', 'the', 'own', 'her', 'few', 'from', 'couldn', 'has', 'and', 'to', 'now', 'or', 't', 'if', \"needn't\", \"hasn't\", 'doing', 'these', 'isn', 'myself', \"couldn't\", 'i', 'more', 'while', 'this', 'at', 'before', 'it', \"won't\", 'any', 'does', 'here', 'further', 'didn', 'having', \"should've\", 'through', 'during', 'haven', \"you'd\", 's', 've', 'weren', \"hadn't\", 'had', 'over', \"didn't\", 'was', 'just', 'because', 'between', 'can', 'we', 'yourselves', \"she's\", 'what', 'should', 'doesn', 'how', 'some', 'ours', 'ma', 'why', 'o', 'once', 'shouldn', 'yours', 'only', 'y', 'those', 'theirs', 'most', 'm', 'too', 'herself', \"isn't\", 'me', 'with', 'hasn', 'mightn', 'wasn', 'your', 'than', 'you'}\n"
     ]
    }
   ],
   "source": [
    "lst = [\"wouldn't\",'won', 'nor', 'not', 'against',\"doesn't\",\"don't\"]\n",
    "for word in lst:\n",
    "    stop.remove(word)\n",
    "print(stop)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sno = nltk.stem.SnowballStemmer('english')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Removal of html tags"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def cleanhtml(sent):\n",
    "    cleanr=re.compile('<.*?>')\n",
    "    cleantext=re.sub(cleanr,' ',sent)\n",
    "    return cleantext"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Removal of puncuation tags"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def cleanpunc(sent):\n",
    "    cleaned=re.sub(r'[?|@|!|^|%|\\'|\"|#|$|:|*|+|-|=|\\-|&|_]',r'',sent)\n",
    "    cleaned=re.sub(r'[.|,|)|(|\\|/]',r' ',cleaned)\n",
    "    return cleaned"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def preprocessing(series):\n",
    "    \n",
    "    i = 0\n",
    "    str1=\" \"\n",
    "    final_string = []    ## This list will contain cleaned sentences\n",
    "    list_of_sent = []    ## This is a list of lists used as input to the W2V model at a later stage\n",
    "    \n",
    "    ## Creating below lists for future use\n",
    "    all_positive_words=[] # store words from spam mail here\n",
    "    all_negative_words=[] # store words from ham mail here\n",
    "    \n",
    "\n",
    "    for sent in series.values:\n",
    "        filtered_sent=[]\n",
    "        sent=cleanhtml(sent)\n",
    "        sent=cleanpunc(sent)\n",
    "        \n",
    "        for clean_word in sent.split():\n",
    "            if((clean_word.isalpha()) and (len(clean_word)>2)):\n",
    "                if(clean_word.lower() not in stop):\n",
    "                    s=(sno.stem(clean_word.lower()))\n",
    "                    s=clean_word.lower()\n",
    "                    filtered_sent.append(s)\n",
    "                    if(data['type'].values)[i]==1:\n",
    "                        all_positive_words.append(s)\n",
    "                    if(data['type'].values)[i]==0:\n",
    "                        all_negative_words.append(s)\n",
    "                        \n",
    "        list_of_sent.append(filtered_sent)\n",
    "        str1=' '.join(filtered_sent)\n",
    "        final_string.append(str1)\n",
    "        i+=1\n",
    "    return final_string,list_of_sent"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "final_string,list_of_sent=preprocessing(data['mail'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(937, 3)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mail</th>\n",
       "      <th>type</th>\n",
       "      <th>cleaned_mail_stem</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>* * * * * * * * * * * * * * * display box cred...</td>\n",
       "      <td>1</td>\n",
       "      <td>display box credit application need place smal...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>dear nlpeople , ' m sure ' ll agree auto insur...</td>\n",
       "      <td>1</td>\n",
       "      <td>dear nlpeople sure agree auto insurance cost m...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>want best hunt camp vacation life , felton 's ...</td>\n",
       "      <td>1</td>\n",
       "      <td>want best hunt camp vacation life felton hunt ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>57 million email addresses $ 99 want money ? p...</td>\n",
       "      <td>1</td>\n",
       "      <td>million email addresses want money put touch m...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>attention ! warn ! adult ! warn ! adult ! unde...</td>\n",
       "      <td>1</td>\n",
       "      <td>attention warn adult warn adult age interest s...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>attention ! warn ! adult ! warn ! adult ! unde...</td>\n",
       "      <td>1</td>\n",
       "      <td>attention warn adult warn adult age interest s...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>subject : re : are debt ? help . qualify finge...</td>\n",
       "      <td>1</td>\n",
       "      <td>subject debt help qualify fingertip long dista...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>= = = = = = = = = = = = = = = = = = = = = = = ...</td>\n",
       "      <td>1</td>\n",
       "      <td>please read found father unite state america w...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>57 million email addresses $ 99 want money ? p...</td>\n",
       "      <td>1</td>\n",
       "      <td>million email addresses want money put touch m...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>$ 3500 per week home computer ! put free softw...</td>\n",
       "      <td>1</td>\n",
       "      <td>per week home computer put free software compu...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                mail  type  \\\n",
       "0  * * * * * * * * * * * * * * * display box cred...     1   \n",
       "1  dear nlpeople , ' m sure ' ll agree auto insur...     1   \n",
       "2  want best hunt camp vacation life , felton 's ...     1   \n",
       "3  57 million email addresses $ 99 want money ? p...     1   \n",
       "4  attention ! warn ! adult ! warn ! adult ! unde...     1   \n",
       "5  attention ! warn ! adult ! warn ! adult ! unde...     1   \n",
       "6  subject : re : are debt ? help . qualify finge...     1   \n",
       "7  = = = = = = = = = = = = = = = = = = = = = = = ...     1   \n",
       "8  57 million email addresses $ 99 want money ? p...     1   \n",
       "9  $ 3500 per week home computer ! put free softw...     1   \n",
       "\n",
       "                                   cleaned_mail_stem  \n",
       "0  display box credit application need place smal...  \n",
       "1  dear nlpeople sure agree auto insurance cost m...  \n",
       "2  want best hunt camp vacation life felton hunt ...  \n",
       "3  million email addresses want money put touch m...  \n",
       "4  attention warn adult warn adult age interest s...  \n",
       "5  attention warn adult warn adult age interest s...  \n",
       "6  subject debt help qualify fingertip long dista...  \n",
       "7  please read found father unite state america w...  \n",
       "8  million email addresses want money put touch m...  \n",
       "9  per week home computer put free software compu...  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['cleaned_mail_stem']=final_string\n",
    "print(data.shape)\n",
    "data.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<function Connection.close>"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conn=sqlite3.connect('spam.sqlite')\n",
    "c=conn.cursor()\n",
    "spam.to_sql('Data',conn,if_exists='replace',index=False)\n",
    "conn.close"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Making Pickle file with cleam_email"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "with open('list_of_sent_for_input_to_w2v.pkl','wb') as pickle_file:\n",
    "    pickle.dump(list_of_sent,pickle_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# BOW"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Shuffling all emails"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mail</th>\n",
       "      <th>type</th>\n",
       "      <th>cleaned_mail_stem</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>791</th>\n",
       "      <td>- - - - - swiss linguistic society organize su...</td>\n",
       "      <td>0</td>\n",
       "      <td>swiss linguistic society organize summer schoo...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>360</th>\n",
       "      <td>hi , where wonder interest subscribe our opt-i...</td>\n",
       "      <td>1</td>\n",
       "      <td>wonder interest subscribe optin email list who...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>582</th>\n",
       "      <td>forward follow post request robert rankin ( ra...</td>\n",
       "      <td>0</td>\n",
       "      <td>forward follow post request robert rankin rank...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>611</th>\n",
       "      <td>* * * * * * celtic linguistics conference * * ...</td>\n",
       "      <td>0</td>\n",
       "      <td>celtic linguistics conference second post call...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>284</th>\n",
       "      <td>discover family history - rate \" cool site wee...</td>\n",
       "      <td>1</td>\n",
       "      <td>discover family history rate cool site week vi...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  mail  type  \\\n",
       "791  - - - - - swiss linguistic society organize su...     0   \n",
       "360  hi , where wonder interest subscribe our opt-i...     1   \n",
       "582  forward follow post request robert rankin ( ra...     0   \n",
       "611  * * * * * * celtic linguistics conference * * ...     0   \n",
       "284  discover family history - rate \" cool site wee...     1   \n",
       "\n",
       "                                     cleaned_mail_stem  \n",
       "791  swiss linguistic society organize summer schoo...  \n",
       "360  wonder interest subscribe optin email list who...  \n",
       "582  forward follow post request robert rankin rank...  \n",
       "611  celtic linguistics conference second post call...  \n",
       "284  discover family history rate cool site week vi...  "
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=shuffle(data,random_state=123)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "bow_vect=CountVectorizer()\n",
    "bow=bow_vect.fit_transform(data['cleaned_mail_stem'].values)\n",
    "x=bow[:,:].toarray()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(937, 22900)"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# tsne"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tsne=TSNE(n_components=2,perplexity=30).fit_transform(x)\n",
    "df=pd.DataFrame(tsne,columns=('dim1','dim2'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0xc259250>"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAFgCAYAAACR2P/oAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x9ff5150>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df=pd.concat([df,data['type'].iloc[:]],axis=1)\n",
    "sns.FacetGrid(df,hue='type',size=5).map(plt.scatter,'dim1','dim2').add_legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Binary - BOW for Bernoulli Naive - Bayes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "bow_vect=CountVectorizer(binary=True)\n",
    "bow=bow_vect.fit_transform(data['cleaned_mail_stem'].values)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x=bow  \n",
    "y=data['type'].values\n",
    "x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "the optimal value of alpha is 0.030\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZgAAAEKCAYAAAAvlUMdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xc1a7cf0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "alpha=np.linspace(0.01,0.1,10)\n",
    "cv_scores=[]\n",
    "\n",
    "for a in alpha:\n",
    "    clf=BernoulliNB(alpha=a)\n",
    "    score=cross_val_score(clf,x_train,y_train,cv=8,scoring='accuracy')\n",
    "    cv_scores.append(score.mean())\n",
    "    \n",
    "MSE=[1-x for x in cv_scores]\n",
    "\n",
    "optimal_alpha=alpha[MSE.index(min(MSE))]\n",
    "print('the optimal value of alpha is %0.3f' % optimal_alpha)\n",
    "plt.plot(alpha,MSE)\n",
    "plt.xlabel('no. of alpha')\n",
    "plt.ylabel('misclassification error')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "99.6453900709\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[145,   1],\n",
       "       [  0, 136]], dtype=int64)"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf=BernoulliNB(alpha=optimal_alpha)\n",
    "clf.fit(x_train,y_train)\n",
    "y_pred=clf.predict(x_test)\n",
    "acc=accuracy_score(y_test,y_pred,normalize=True)*float(100)\n",
    "print(acc)\n",
    "cof=confusion_matrix(y_test,y_pred).T\n",
    "cof"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision is = 1.0\n",
      "recall is = 0.9927007299270073\n",
      "fscore is = 0.9963369963369962\n"
     ]
    }
   ],
   "source": [
    "tp=int(cof[1:,1:])\n",
    "fp=int(cof[1:,:1])\n",
    "fn=int(cof[:1,1:])\n",
    "pre=tp/(tp+fp)\n",
    "rec=tp/(tp+fn)\n",
    "fsco=(2*pre*rec)/(pre+rec)\n",
    "print(\"precision is =\",pre)\n",
    "print(\"recall is =\",rec)\n",
    "print(\"fscore is =\",fsco)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# multinomailNB - BOW"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "bow_vect=CountVectorizer()\n",
    "bow=bow_vect.fit_transform(data['cleaned_mail_stem'].values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x=bow  \n",
    "y=data['type'].values\n",
    "x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "the optimal value of alpha is 0.043\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZgAAAEKCAYAAAAvlUMdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nPBSfXwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xd226d10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "alpha=np.linspace(0.01,0.1,20)\n",
    "cv_scores=[]\n",
    "\n",
    "for a in alpha:\n",
    "    clf=MultinomialNB(alpha=a)\n",
    "    score=cross_val_score(clf,x_train,y_train,cv=8,scoring='accuracy')\n",
    "    cv_scores.append(score.mean())\n",
    "    \n",
    "MSE=[1-x for x in cv_scores]\n",
    "\n",
    "optimal_alpha=alpha[MSE.index(min(MSE))]\n",
    "print('the optimal value of alpha is %0.3f' % optimal_alpha)\n",
    "plt.plot(alpha,MSE)\n",
    "plt.xlabel('no. of alpha')\n",
    "plt.ylabel('misclassification error')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "98.9361702128\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[145,   3],\n",
       "       [  0, 134]], dtype=int64)"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf=MultinomialNB(alpha=optimal_alpha)\n",
    "clf.fit(x_train,y_train)\n",
    "y_pred=clf.predict(x_test)\n",
    "acc=accuracy_score(y_test,y_pred,normalize=True)*float(100)\n",
    "print(acc)\n",
    "cof=confusion_matrix(y_test,y_pred).T\n",
    "cof"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision is = 1.0\n",
      "recall is = 0.9781021897810219\n",
      "fscore is = 0.988929889298893\n"
     ]
    }
   ],
   "source": [
    "tp=int(cof[1:,1:])\n",
    "fp=int(cof[1:,:1])\n",
    "fn=int(cof[:1,1:])\n",
    "pre=tp/(tp+fp)\n",
    "rec=tp/(tp+fn)\n",
    "fsco=(2*pre*rec)/(pre+rec)\n",
    "print(\"precision is =\",pre)\n",
    "print(\"recall is =\",rec)\n",
    "print(\"fscore is =\",fsco)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# tdif tsne"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tfidf_vec=TfidfVectorizer()\n",
    "tfidf=tfidf_vec.fit_transform(data['cleaned_mail_stem'].values)\n",
    "y=tfidf[:,:].toarray()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0xc2694d0>"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAFgCAYAAACR2P/oAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nKisaSJRSSmVFA4lSSqms/H899iNWBOelEQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xd05d3d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tsne=TSNE(n_components=2,perplexity=30).fit_transform(y)\n",
    "df=pd.DataFrame(tsne,columns=('dim1','dim2'))\n",
    "df=pd.concat([df,data['type'].iloc[:3500]],axis=1)\n",
    "sns.FacetGrid(df,hue='type',size=5).map(plt.scatter,'dim1','dim2').add_legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# tdif -multinomailNB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tf_idf=TfidfVectorizer()\n",
    "tfidf=tf_idf.fit_transform(data['cleaned_mail_stem'].values)\n",
    "x=tfidf  \n",
    "y=data['type'].values\n",
    "x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "the optimal value of alpha is 0.43\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZgAAAEKCAYAAAAvlUMdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nxP8PUWLYPqTlzrgAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xc710ad0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "alpha=np.linspace(0.1,1,20)\n",
    "cv_scores=[]\n",
    "\n",
    "for a in alpha:\n",
    "    clf=MultinomialNB(alpha=a)\n",
    "    score=cross_val_score(clf,x_train,y_train,cv=8,scoring='accuracy')\n",
    "    cv_scores.append(score.mean())\n",
    "    \n",
    "MSE=[1-x for x in cv_scores]\n",
    "\n",
    "optimal_alpha=alpha[MSE.index(min(MSE))]\n",
    "print('the optimal value of alpha is %0.2f' % optimal_alpha)\n",
    "plt.plot(alpha,MSE)\n",
    "plt.xlabel('no. of alpha')\n",
    "plt.ylabel('misclassification error')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "98.5815602837\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[142,   1],\n",
       "       [  3, 136]], dtype=int64)"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf=MultinomialNB(alpha=optimal_alpha)\n",
    "clf.fit(x_train,y_train)\n",
    "y_pred=clf.predict(x_test)\n",
    "acc=accuracy_score(y_test,y_pred,normalize=True)*float(100)\n",
    "print(acc)\n",
    "cof=confusion_matrix(y_test,y_pred).T\n",
    "cof"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision is = 0.9784172661870504\n",
      "recall is = 0.9927007299270073\n",
      "fscore is = 0.9855072463768116\n"
     ]
    }
   ],
   "source": [
    "tp=int(cof[1:,1:])\n",
    "fp=int(cof[1:,:1])\n",
    "fn=int(cof[:1,1:])\n",
    "pre=tp/(tp+fp)\n",
    "rec=tp/(tp+fn)\n",
    "fsco=(2*pre*rec)/(pre+rec)\n",
    "print(\"precision is =\",pre)\n",
    "print(\"recall is =\",rec)\n",
    "print(\"fscore is =\",fsco)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CONCLUSION\n",
    "* Amongst all model Binary BOW has excellent confusion matrix with correct prediction and also high acc,recall,fscore\n",
    "* we have tdif is showing more overlap but still its showing goood result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}