model 2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "documented-costa",
   "metadata": {},
   "source": [
    "### Embedding With CNN Model For Sentiment Analysis\n",
    "**Embedding**\n",
    "    - Word embeddings are a technique for representing text where different words with similar meaning have a similar real-valued vector representation.\n",
    "------------------------------------------------------------------------------\n",
    "\n",
    "**Workflow**\n",
    " \n",
    " - Dataset\n",
    " - Data Prepration\n",
    " - Train CNN with Embedding Layer\n",
    " - Evaluate Model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "agricultural-lodge",
   "metadata": {},
   "source": [
    "**Dataset**\n",
    "https://www.kaggle.com/kazanova/sentiment140"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "id": "defensive-nursing",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 227,
   "id": "matched-natural",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('./dataset/tweets.csv', header = None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 228,
   "id": "basic-vessel",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1600000, 6)"
      ]
     },
     "execution_count": 228,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 229,
   "id": "floppy-cursor",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810369</td>\n",
       "      <td>Mon Apr 06 22:19:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_TheSpecialOne_</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810672</td>\n",
       "      <td>Mon Apr 06 22:19:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>scotthamilton</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810917</td>\n",
       "      <td>Mon Apr 06 22:19:53 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mattycus</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811184</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ElleCTF</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811193</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Karoli</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   0           1                             2         3                4  \\\n",
       "0  0  1467810369  Mon Apr 06 22:19:45 PDT 2009  NO_QUERY  _TheSpecialOne_   \n",
       "1  0  1467810672  Mon Apr 06 22:19:49 PDT 2009  NO_QUERY    scotthamilton   \n",
       "2  0  1467810917  Mon Apr 06 22:19:53 PDT 2009  NO_QUERY         mattycus   \n",
       "3  0  1467811184  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY          ElleCTF   \n",
       "4  0  1467811193  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY           Karoli   \n",
       "\n",
       "                                                   5  \n",
       "0  @switchfoot http://twitpic.com/2y1zl - Awww, t...  \n",
       "1  is upset that he can't update his Facebook by ...  \n",
       "2  @Kenichan I dived many times for the ball. Man...  \n",
       "3    my whole body feels itchy and like its on fire   \n",
       "4  @nationwideclass no, it's not behaving at all....  "
      ]
     },
     "execution_count": 229,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 230,
   "id": "steady-birthday",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.rename(columns={0: 'target', 1: 'id', 2: 'date', 3: 'query', 4: 'username', 5: 'content'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 231,
   "id": "bound-intranet",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>id</th>\n",
       "      <th>date</th>\n",
       "      <th>query</th>\n",
       "      <th>username</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810369</td>\n",
       "      <td>Mon Apr 06 22:19:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_TheSpecialOne_</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810672</td>\n",
       "      <td>Mon Apr 06 22:19:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>scotthamilton</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810917</td>\n",
       "      <td>Mon Apr 06 22:19:53 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mattycus</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811184</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ElleCTF</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811193</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Karoli</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   target          id                          date     query  \\\n",
       "0       0  1467810369  Mon Apr 06 22:19:45 PDT 2009  NO_QUERY   \n",
       "1       0  1467810672  Mon Apr 06 22:19:49 PDT 2009  NO_QUERY   \n",
       "2       0  1467810917  Mon Apr 06 22:19:53 PDT 2009  NO_QUERY   \n",
       "3       0  1467811184  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "4       0  1467811193  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "\n",
       "          username                                            content  \n",
       "0  _TheSpecialOne_  @switchfoot http://twitpic.com/2y1zl - Awww, t...  \n",
       "1    scotthamilton  is upset that he can't update his Facebook by ...  \n",
       "2         mattycus  @Kenichan I dived many times for the ball. Man...  \n",
       "3          ElleCTF    my whole body feels itchy and like its on fire   \n",
       "4           Karoli  @nationwideclass no, it's not behaving at all....  "
      ]
     },
     "execution_count": 231,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 232,
   "id": "placed-thesaurus",
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.set_option('display.max_columns', 2000)\n",
    "pd.set_option('display.max_rows', 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 233,
   "id": "approximate-mouse",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "800000\n",
      "0\n",
      "800000\n"
     ]
    }
   ],
   "source": [
    "negativeTweets = df[df['target']==0]['content'].count()\n",
    "print(negativeTweets)\n",
    "neutralTweets = df[df['target']==2]['content'].count()\n",
    "print(neutralTweets)\n",
    "positiveTweets = df[df['target']==4]['content'].count()\n",
    "print(positiveTweets)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 234,
   "id": "bigger-destiny",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['target'] = df['target'].replace([0,4],[0,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 235,
   "id": "original-single",
   "metadata": {},
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 236,
   "id": "stylish-forwarding",
   "metadata": {},
   "outputs": [],
   "source": [
    "# plt.figure(figsize=(14,7))\n",
    "# common_keyword=sns.barplot(x=df[df['target']==1]['username'].value_counts()[:10].index,\n",
    "#                            y=df[df['target']==1]['username'].value_counts()[:10],palette='magma')\n",
    "# common_keyword.set_xticklabels(common_keyword.get_xticklabels(),rotation=90)\n",
    "# common_keyword.set_ylabel('Positive tweet frequency',fontsize=12)\n",
    "# plt.title('Top 10 users who publish positive tweets',fontsize=16)\n",
    "# plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "id": "forbidden-wrapping",
   "metadata": {},
   "outputs": [],
   "source": [
    "# plt.figure(figsize=(14,7))\n",
    "# common_keyword=sns.barplot(x=df[df['target']==0]['username'].value_counts()[:10].index,\n",
    "#                            y=df[df['target']==0]['username'].value_counts()[:10],palette='magma')\n",
    "# common_keyword.set_xticklabels(common_keyword.get_xticklabels(),rotation=90)\n",
    "# common_keyword.set_ylabel('Positive tweet frequency',fontsize=12)\n",
    "# plt.title('Top 10 users who publish positive tweets',fontsize=16)\n",
    "# plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "id": "iraqi-possibility",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.drop(['id','date','query','username'], axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "id": "subsequent-serbia",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   target                                            content\n",
       "0       0  @switchfoot http://twitpic.com/2y1zl - Awww, t...\n",
       "1       0  is upset that he can't update his Facebook by ...\n",
       "2       0  @Kenichan I dived many times for the ball. Man...\n",
       "3       0    my whole body feels itchy and like its on fire \n",
       "4       0  @nationwideclass no, it's not behaving at all...."
      ]
     },
     "execution_count": 239,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "hindu-iraqi",
   "metadata": {},
   "source": [
    "### Data prepration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 240,
   "id": "appropriate-couple",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.utils import shuffle\n",
    "df = shuffle(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "id": "bound-supplier",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.iloc[0:100000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 242,
   "id": "revised-hanging",
   "metadata": {
    "scrolled": true
   },
   "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>target</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1148909</th>\n",
       "      <td>1</td>\n",
       "      <td>I am really bad at this Twitter thing, but if ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5214</th>\n",
       "      <td>0</td>\n",
       "      <td>@ChrisInBoston Baston... that sucks dude.    {...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>401271</th>\n",
       "      <td>0</td>\n",
       "      <td>holy shit its already 4!! omfg what a waste of...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1514675</th>\n",
       "      <td>1</td>\n",
       "      <td>Meg joined the Pig of the Month competion  htt...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1243242</th>\n",
       "      <td>1</td>\n",
       "      <td>@Azikale Are you suffering with Dave Syndrome?...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1356301</th>\n",
       "      <td>1</td>\n",
       "      <td>@Katashtrophy I'm not a gamer but I am a BOARD...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1218437</th>\n",
       "      <td>1</td>\n",
       "      <td>@SarahJacinta Thank you!</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>761592</th>\n",
       "      <td>0</td>\n",
       "      <td>Just finished a bag of Fritos. My head hurts</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>112455</th>\n",
       "      <td>0</td>\n",
       "      <td>@RespectMileyC You have school tomorrow   aw t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>254181</th>\n",
       "      <td>0</td>\n",
       "      <td>I has no juice and i just vaccumed up a pink p...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>630654</th>\n",
       "      <td>0</td>\n",
       "      <td>@ddlovato, @mileycyrus, @taylorswift, and @sel...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>545224</th>\n",
       "      <td>0</td>\n",
       "      <td>some people in this world are just so sick. it...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1463532</th>\n",
       "      <td>1</td>\n",
       "      <td>@Katersoneseven honey and lemon in boiling wat...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1465873</th>\n",
       "      <td>1</td>\n",
       "      <td>@f4tty Received loud and clear! I'll send you ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>399054</th>\n",
       "      <td>0</td>\n",
       "      <td>is sick  with a cold....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>224076</th>\n",
       "      <td>0</td>\n",
       "      <td>@JuniesgurlIt's the last night for me to be on...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>557506</th>\n",
       "      <td>0</td>\n",
       "      <td>@chrys73 IM HAPPY 4 Y YES YES YES!! IM ANNOYED...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1102102</th>\n",
       "      <td>1</td>\n",
       "      <td>@alexanderrybak jeg elsker deg!!! (:  i got my...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>648874</th>\n",
       "      <td>0</td>\n",
       "      <td>@bethg40   hope it starts to feeling like Friday!</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>869307</th>\n",
       "      <td>1</td>\n",
       "      <td>@thomasnesse @plo @nrkbeta SVT (  http://bit.l...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>378266</th>\n",
       "      <td>0</td>\n",
       "      <td>iTunes messes up the order of the songs in an ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1245886</th>\n",
       "      <td>1</td>\n",
       "      <td>@MichelleCarrick i wasn't expecting it so yeah...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>976534</th>\n",
       "      <td>1</td>\n",
       "      <td>is now uploading vids on mulitply.  http://plu...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>733629</th>\n",
       "      <td>0</td>\n",
       "      <td>Has a very bad hangover, Also how much does th...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>944758</th>\n",
       "      <td>1</td>\n",
       "      <td>wasssup everyone - weekend is near</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1161832</th>\n",
       "      <td>1</td>\n",
       "      <td>@tombarrett got it! thank you...still am confu...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1170514</th>\n",
       "      <td>1</td>\n",
       "      <td>@MinBitt Depends on the child, but I would thi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1504739</th>\n",
       "      <td>1</td>\n",
       "      <td>@GabrielMY Great cant wait to read them</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>467269</th>\n",
       "      <td>0</td>\n",
       "      <td>@ayatoshirosan ...received the mailgram of whi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1466678</th>\n",
       "      <td>1</td>\n",
       "      <td>@ave77 be sure to visit http://www.MedanKu.com</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         target                                            content\n",
       "1148909       1  I am really bad at this Twitter thing, but if ...\n",
       "5214          0  @ChrisInBoston Baston... that sucks dude.    {...\n",
       "401271        0  holy shit its already 4!! omfg what a waste of...\n",
       "1514675       1  Meg joined the Pig of the Month competion  htt...\n",
       "1243242       1  @Azikale Are you suffering with Dave Syndrome?...\n",
       "1356301       1  @Katashtrophy I'm not a gamer but I am a BOARD...\n",
       "1218437       1                          @SarahJacinta Thank you! \n",
       "761592        0      Just finished a bag of Fritos. My head hurts \n",
       "112455        0  @RespectMileyC You have school tomorrow   aw t...\n",
       "254181        0  I has no juice and i just vaccumed up a pink p...\n",
       "630654        0  @ddlovato, @mileycyrus, @taylorswift, and @sel...\n",
       "545224        0  some people in this world are just so sick. it...\n",
       "1463532       1  @Katersoneseven honey and lemon in boiling wat...\n",
       "1465873       1  @f4tty Received loud and clear! I'll send you ...\n",
       "399054        0                          is sick  with a cold.... \n",
       "224076        0  @JuniesgurlIt's the last night for me to be on...\n",
       "557506        0  @chrys73 IM HAPPY 4 Y YES YES YES!! IM ANNOYED...\n",
       "1102102       1  @alexanderrybak jeg elsker deg!!! (:  i got my...\n",
       "648874        0  @bethg40   hope it starts to feeling like Friday!\n",
       "869307        1  @thomasnesse @plo @nrkbeta SVT (  http://bit.l...\n",
       "378266        0  iTunes messes up the order of the songs in an ...\n",
       "1245886       1  @MichelleCarrick i wasn't expecting it so yeah...\n",
       "976534        1  is now uploading vids on mulitply.  http://plu...\n",
       "733629        0  Has a very bad hangover, Also how much does th...\n",
       "944758        1                wasssup everyone - weekend is near \n",
       "1161832       1  @tombarrett got it! thank you...still am confu...\n",
       "1170514       1  @MinBitt Depends on the child, but I would thi...\n",
       "1504739       1           @GabrielMY Great cant wait to read them \n",
       "467269        0  @ayatoshirosan ...received the mailgram of whi...\n",
       "1466678       1    @ave77 be sure to visit http://www.MedanKu.com "
      ]
     },
     "execution_count": 242,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 243,
   "id": "industrial-sewing",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50020, 49980)"
      ]
     },
     "execution_count": 243,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df['target']==0]['target'].count(),df[df['target']==1]['target'].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 244,
   "id": "typical-owner",
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.corpus import stopwords\n",
    "import string\n",
    "import re\n",
    "from nltk.stem.snowball import SnowballStemmer "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 245,
   "id": "sufficient-argument",
   "metadata": {},
   "outputs": [],
   "source": [
    "regex = \"@\\S+|https?:\\S+|http?:\\S|[^A-Za-z0-9]+\" \n",
    "english_stopwords = stopwords.words('english')\n",
    "stemmer = SnowballStemmer('english')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 246,
   "id": "significant-replacement",
   "metadata": {},
   "outputs": [],
   "source": [
    "def clean_data(content, stem=False):\n",
    "    content = re.sub(regex, ' ', str(content).lower()).strip()\n",
    "    tokens = []\n",
    "    for token in content.split():\n",
    "        if token not in english_stopwords:\n",
    "              tokens.append(stemmer.stem(token))\n",
    "    return \" \".join(tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 247,
   "id": "concrete-disability",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.content = df.content.apply(lambda x: clean_data(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 248,
   "id": "moving-quantum",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1148909</th>\n",
       "      <td>1</td>\n",
       "      <td>realli bad twitter thing anyon would like get ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5214</th>\n",
       "      <td>0</td>\n",
       "      <td>baston suck dude send happi thought way</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>401271</th>\n",
       "      <td>0</td>\n",
       "      <td>holi shit alreadi 4 omfg wast saturday</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1514675</th>\n",
       "      <td>1</td>\n",
       "      <td>meg join pig month compet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1243242</th>\n",
       "      <td>1</td>\n",
       "      <td>suffer dave syndrom jump roof car loincloth</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         target                                            content\n",
       "1148909       1  realli bad twitter thing anyon would like get ...\n",
       "5214          0            baston suck dude send happi thought way\n",
       "401271        0             holi shit alreadi 4 omfg wast saturday\n",
       "1514675       1                          meg join pig month compet\n",
       "1243242       1        suffer dave syndrom jump roof car loincloth"
      ]
     },
     "execution_count": 248,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 249,
   "id": "grand-investment",
   "metadata": {},
   "outputs": [],
   "source": [
    "#df.to_csv('cleadDataset.csv', header=False, index=False) "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "innocent-passion",
   "metadata": {},
   "source": [
    "### Define A Vocabulary\n",
    "- We can develop a vocabulary as a Counter, which is a dictionary mapping of words and their count that allows us to easily update and query. Each document can be added to the counter (a new function called add doc to vocab()) and we can step over all of the tweets in the negative directory and then the positive directory (a new function called process docs()).\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 250,
   "id": "through-symbol",
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import Counter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 251,
   "id": "initial-export",
   "metadata": {},
   "outputs": [],
   "source": [
    "def clean_tweet(doc):\n",
    "    tokens = doc.split()\n",
    "    stop_words = set(stopwords.words('english'))\n",
    "    tokens = [w for w in tokens if not w in stop_words and len(w) > 1]\n",
    "    return tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 252,
   "id": "challenging-memory",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load tweets and add to vocab/dictionary\n",
    "def add_to_vocab(tweets, vocab):\n",
    "    #load tweet\n",
    "    for x in tweets.content:\n",
    "        tokens = clean_tweet(x)\n",
    "        vocab.update(tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 253,
   "id": "expressed-meeting",
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab = Counter()\n",
    "add_to_vocab(df, vocab)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 254,
   "id": "affected-electric",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "47125"
      ]
     },
     "execution_count": 254,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(vocab)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 255,
   "id": "elegant-business",
   "metadata": {},
   "outputs": [],
   "source": [
    "# vocab.items()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 256,
   "id": "affected-directive",
   "metadata": {},
   "outputs": [],
   "source": [
    "# import json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 257,
   "id": "conventional-myrtle",
   "metadata": {},
   "outputs": [],
   "source": [
    "# vocabToJson = json.dumps(vocab, indent = 4)   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 258,
   "id": "indoor-implement",
   "metadata": {},
   "outputs": [],
   "source": [
    "# with open('dictionary.json', 'w+') as f:\n",
    "#     # this would place the entire output on one line\n",
    "#     # use json.dump(lista_items, f, indent=4) to \"pretty-print\" with four spaces per indent\n",
    "#     json.dump(vocabToJson, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 259,
   "id": "capable-fundamentals",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "17695\n"
     ]
    }
   ],
   "source": [
    "#keep tokens with min occurences\n",
    "min_occurance = 2\n",
    "tokens = [k for k, c in vocab.items() if c >= min_occurance]\n",
    "print(len(tokens))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 260,
   "id": "stunning-release",
   "metadata": {},
   "outputs": [],
   "source": [
    "# save list to file\n",
    "def save_list(lines, filename):\n",
    "    #convert lines to a single blob of text\n",
    "    data = '\\n'.join(lines)\n",
    "    #open file \n",
    "    file = open(filename,'w')\n",
    "    #write text\n",
    "    file.write(data)\n",
    "    #close file\n",
    "    file.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 261,
   "id": "prostate-confidentiality",
   "metadata": {},
   "outputs": [],
   "source": [
    "#save_list(tokens,'filteredDictionary.txt')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "comfortable-intro",
   "metadata": {},
   "source": [
    "### Train CNN with Embedding Layer\n",
    "\n",
    "A word embedding is a way of representing text where each word in\n",
    "the vocabulary is represented by a real valued vector in a high-dimensional space. The vectors\n",
    "are learned in such a way that words that have similar meanings will have similar representation\n",
    "in the vector space (close in the vector space). This is a more expressive representation for text\n",
    "than more classical methods like bag-of-words, where relationships between words or tokens are\n",
    "ignored, or forced in bigram and trigram approaches.\n",
    "\n",
    "\n",
    "The real valued vector representation for words can be learned while training the neural\n",
    "network. We can do this in the Keras deep learning library using the Embedding layer. The\n",
    "first step is to load the vocabulary. We will use it to filter out words from tweets that\n",
    "we are not interested in."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 262,
   "id": "banned-independence",
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_doc(filename):\n",
    "    # open the file as read only\n",
    "    file = open(filename, 'r')\n",
    "    text = file.read()\n",
    "    file.close()\n",
    "    return text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 263,
   "id": "smoking-guess",
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab_file = 'filteredDictionary.txt'\n",
    "vocab = load_doc(vocab_file)\n",
    "vocab = set(vocab.split())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 264,
   "id": "honest-interpretation",
   "metadata": {},
   "outputs": [],
   "source": [
    "# fit a tokenizer\n",
    "def create_tokenizer(lines):\n",
    "    tokenizer = Tokenizer()\n",
    "    tokenizer.fit_on_texts(lines)\n",
    "    return tokenizer\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 265,
   "id": "sublime-chase",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum length: 34\n"
     ]
    }
   ],
   "source": [
    "max_length = max([len(s.split()) for s in df.content])\n",
    "print('Maximum length: %d' % max_length)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 266,
   "id": "southeast-loading",
   "metadata": {},
   "outputs": [],
   "source": [
    "# integer encode and pad documents\n",
    "def encode_docs(tokenizer, max_length, docs):\n",
    "    # integer encode\n",
    "    encoded = tokenizer.texts_to_sequences(docs)\n",
    "    # pad sequences\n",
    "    padded = pad_sequences(encoded, maxlen=max_length, padding='post')\n",
    "    return padded\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 267,
   "id": "applicable-cream",
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing.text import Tokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 268,
   "id": "seeing-crown",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = Tokenizer()\n",
    "tokenizer.fit_on_texts(df.content)\n",
    "vocab_size = len(tokenizer.word_index) + 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 269,
   "id": "tribal-theory",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary size: 47204\n"
     ]
    }
   ],
   "source": [
    "# define vocabulary size\n",
    "print('Vocabulary size: %d' % vocab_size)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 270,
   "id": "collaborative-laser",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 271,
   "id": "pacific-vegetation",
   "metadata": {},
   "outputs": [],
   "source": [
    "train, test = train_test_split(df, test_size = 0.3, random_state = 4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 272,
   "id": "passive-graham",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train dataset size: (70000, 2)\n",
      "test dataset size: (30000, 2)\n"
     ]
    }
   ],
   "source": [
    "print('train dataset size: {}'.format(train.shape))\n",
    "print('test dataset size: {}'.format(test.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 273,
   "id": "serious-relief",
   "metadata": {},
   "outputs": [],
   "source": [
    "train = shuffle(train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 274,
   "id": "killing-director",
   "metadata": {},
   "outputs": [],
   "source": [
    "test = shuffle(test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "other-diesel",
   "metadata": {},
   "source": [
    "### Define Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 275,
   "id": "tracked-apache",
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import seaborn as sns\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Activation, Dense, Dropout, Embedding, Flatten, Conv1D, MaxPooling1D, LSTM\n",
    "from keras import utils\n",
    "from keras.callbacks import ReduceLROnPlateau, EarlyStopping\n",
    "from keras.optimizers import Adam"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 276,
   "id": "jewish-constitution",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_sequences = tokenizer.texts_to_sequences(train.content)\n",
    "test_sequences = tokenizer.texts_to_sequences(test.content)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 277,
   "id": "killing-tribute",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train =  pad_sequences(train_sequences, maxlen = max_length, padding = 'post')\n",
    "X_test = pad_sequences(test_sequences, maxlen = max_length, padding = 'post')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 294,
   "id": "abandoned-geneva",
   "metadata": {},
   "outputs": [],
   "source": [
    "def encode_tweet(tokenizer, max_length, tweet):\n",
    "    encoded = tokenizer.texts_to_sequences(tweet)\n",
    "    padded = pad_sequences(train_sequences, maxlen = max_length, padding = 'post')\n",
    "    return padded"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 278,
   "id": "intermediate-virgin",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  63,  258,    0, ...,    0,    0,    0],\n",
       "       [ 380,   25, 1447, ...,    0,    0,    0],\n",
       "       [ 113,   36,    0, ...,    0,    0,    0],\n",
       "       ...,\n",
       "       [   1,   49,    0, ...,    0,    0,    0],\n",
       "       [ 777,  826,   74, ...,    0,    0,    0],\n",
       "       [   7,   16,   42, ...,    0,    0,    0]])"
      ]
     },
     "execution_count": 278,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 279,
   "id": "returning-winner",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(70000, 34)"
      ]
     },
     "execution_count": 279,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 280,
   "id": "consistent-result",
   "metadata": {},
   "outputs": [],
   "source": [
    "Y_train = train.target.values # np array\n",
    "Y_test = test.target.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 281,
   "id": "piano-subcommittee",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 0, 1, 1, 1, 1, 0, 0, 0, 1], dtype=int64)"
      ]
     },
     "execution_count": 281,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_train[0:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 282,
   "id": "local-tiger",
   "metadata": {},
   "outputs": [],
   "source": [
    "def define_model(vocab_size, max_length):\n",
    "    model = Sequential()\n",
    "    model.add(Embedding(vocab_size, 100, input_length=max_length))\n",
    "    model.add(Conv1D(filters=32, kernel_size=8, activation='relu'))\n",
    "    model.add(MaxPooling1D(pool_size=2))\n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(10, activation='relu'))\n",
    "    model.add(Dense(1, activation='sigmoid'))\n",
    "    # compile network\n",
    "    model.compile(loss='binary_crossentropy', optimizer = Adam(learning_rate=0.001), metrics=['accuracy'])\n",
    "    # summarize defined model\n",
    "    model.summary()\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 292,
   "id": "instant-correction",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_4\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_4 (Embedding)      (None, 34, 100)           4720400   \n",
      "_________________________________________________________________\n",
      "conv1d_4 (Conv1D)            (None, 27, 32)            25632     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_4 (MaxPooling1 (None, 13, 32)            0         \n",
      "_________________________________________________________________\n",
      "flatten_4 (Flatten)          (None, 416)               0         \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 10)                4170      \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 1)                 11        \n",
      "=================================================================\n",
      "Total params: 4,750,213\n",
      "Trainable params: 4,750,213\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = define_model(vocab_size, max_length)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 293,
   "id": "induced-charter",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 293,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pydot\n",
    "import graphviz\n",
    "tf.keras.utils.plot_model(model, show_shapes=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 285,
   "id": "general-evaluation",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "274/274 - 25s - loss: 0.5730 - accuracy: 0.6854 - val_loss: 0.4990 - val_accuracy: 0.7583\n",
      "Epoch 2/10\n",
      "274/274 - 22s - loss: 0.4290 - accuracy: 0.8070 - val_loss: 0.5077 - val_accuracy: 0.7558\n",
      "Epoch 3/10\n",
      "274/274 - 22s - loss: 0.3175 - accuracy: 0.8673 - val_loss: 0.5645 - val_accuracy: 0.7424\n",
      "Epoch 4/10\n",
      "274/274 - 24s - loss: 0.2138 - accuracy: 0.9161 - val_loss: 0.6754 - val_accuracy: 0.7292\n",
      "Epoch 5/10\n",
      "274/274 - 23s - loss: 0.1524 - accuracy: 0.9417 - val_loss: 0.7791 - val_accuracy: 0.7180\n",
      "Epoch 6/10\n",
      "274/274 - 23s - loss: 0.1154 - accuracy: 0.9569 - val_loss: 0.8915 - val_accuracy: 0.7142\n",
      "Epoch 7/10\n",
      "274/274 - 23s - loss: 0.0926 - accuracy: 0.9654 - val_loss: 0.9873 - val_accuracy: 0.7068\n",
      "Epoch 8/10\n",
      "274/274 - 22s - loss: 0.0768 - accuracy: 0.9709 - val_loss: 1.0947 - val_accuracy: 0.7045\n",
      "Epoch 9/10\n",
      "274/274 - 22s - loss: 0.0646 - accuracy: 0.9755 - val_loss: 1.2009 - val_accuracy: 0.7018\n",
      "Epoch 10/10\n",
      "274/274 - 26s - loss: 0.0552 - accuracy: 0.9790 - val_loss: 1.3658 - val_accuracy: 0.6972\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, Y_train, batch_size = 256, epochs=10, validation_data=(X_test, Y_test), verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 286,
   "id": "partial-sense",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "y_pred = model.predict(X_test)\n",
    "y_pred = np.where(y_pred>0.5, 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 287,
   "id": "applicable-breed",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.70      0.69      0.70     15123\n",
      "           1       0.69      0.70      0.70     14877\n",
      "\n",
      "    accuracy                           0.70     30000\n",
      "   macro avg       0.70      0.70      0.70     30000\n",
      "weighted avg       0.70      0.70      0.70     30000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(Y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 288,
   "id": "sufficient-making",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 289,
   "id": "plain-guide",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuracy : 98.351431 \n"
     ]
    }
   ],
   "source": [
    "_, acc = model.evaluate(X_train, Y_train, verbose = 0)\n",
    "print('Training Accuracy : %f ' %(acc*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 290,
   "id": "ceramic-active",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Testing Accuracy : 69.716668 \n"
     ]
    }
   ],
   "source": [
    "_, acc = model.evaluate(X_test, Y_test, verbose = 0)\n",
    "print('Testing Accuracy : %f ' %(acc*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 291,
   "id": "surprising-ranking",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#History for accuracy\n",
    "plt.figure(figsize=(10,5))\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['Train accuracy', 'Test accuracy'], loc='lower right')\n",
    "plt.show()\n",
    "# History for loss\n",
    "plt.figure(figsize=(10,5))\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['Train loss', 'Test loss'], loc='upper right')\n",
    "plt.suptitle('Accuracy and loss for second model')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 356,
   "id": "neither-processor",
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict_sentiment(review, vocab, tokenizer, max_length, model):\n",
    "    line = clean_data(review, stem = True)\n",
    "    padded = encode_tweet(tokenizer,max_length, review)\n",
    "    yhat = model.predict(padded, verbose = 0)\n",
    "    percent_pos = yhat[0,0]\n",
    "    if round(percent_pos) == 0:\n",
    "        return (1-percent_pos), \"Negative\"\n",
    "    return percent_pos, 'Positive'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 357,
   "id": "nasty-music",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Volume in drive C is OS\n",
      " Volume Serial Number is 8ADB-680F\n",
      "\n",
      " Directory of C:\\Users\\dell\\Desktop\\Machine Learning\\Project Twitter Sentiment Analysis\n",
      "\n",
      "17-03-2021  21:48    <DIR>          .\n",
      "17-03-2021  21:48    <DIR>          ..\n",
      "17-03-2021  12:05    <DIR>          .ipynb_checkpoints\n",
      "16-03-2021  20:35               355 app.py\n",
      "17-03-2021  12:51        68,440,533 cleadDataset.csv\n",
      "17-03-2021  12:29    <DIR>          dataset\n",
      "17-03-2021  13:45         6,212,519 dictionary.json\n",
      "12-03-2021  13:50             8,901 download.jpg\n",
      "17-03-2021  13:50           840,869 filteredDictionary.txt\n",
      "14-06-2020  18:44       347,116,733 glove.6B.100d.txt\n",
      "12-03-2021  17:27       137,847,651 glove100dimentional.zip\n",
      "17-03-2021  19:46            32,055 model.png\n",
      "12-03-2021  22:32                39 README.md\n",
      "17-03-2021  15:23           122,411 Twitter Sentiment Analysis.ipynb\n",
      "12-03-2021  22:09            11,149 Twitter Sentiment Analysis.py\n",
      "17-03-2021  21:48           150,511 Untitled.ipynb\n",
      "              12 File(s)    560,783,726 bytes\n",
      "               4 Dir(s)  745,201,254,400 bytes free\n"
     ]
    }
   ],
   "source": [
    "!dir"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 358,
   "id": "editorial-replacement",
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab_filename = 'filteredDictionary.txt'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 359,
   "id": "casual-typing",
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab = load_doc(vocab_filename)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 360,
   "id": "pleased-mason",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "pandas.core.frame.DataFrame"
      ]
     },
     "execution_count": 360,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 361,
   "id": "dying-special",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(70000, 2)"
      ]
     },
     "execution_count": 361,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 362,
   "id": "human-poker",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>609352</th>\n",
       "      <td>0</td>\n",
       "      <td>sorri anyth</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>294026</th>\n",
       "      <td>0</td>\n",
       "      <td>heart amp prayer go passeng amp famili air fra...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1037682</th>\n",
       "      <td>1</td>\n",
       "      <td>alway look</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1344115</th>\n",
       "      <td>1</td>\n",
       "      <td>loveeee kobe highest trend topic kobe kobe kobe</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1293327</th>\n",
       "      <td>1</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         target                                            content\n",
       "609352        0                                        sorri anyth\n",
       "294026        0  heart amp prayer go passeng amp famili air fra...\n",
       "1037682       1                                         alway look\n",
       "1344115       1    loveeee kobe highest trend topic kobe kobe kobe\n",
       "1293327       1                                                   "
      ]
     },
     "execution_count": 362,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 363,
   "id": "charitable-village",
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_tokenizer(train):\n",
    "    tokenizer = Tokenizer()\n",
    "    tokenizer.fit_on_texts(train.content)\n",
    "#     vocab_size = len(tokenizer.word_index) + 1\n",
    "    return tokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 364,
   "id": "variable-germany",
   "metadata": {},
   "outputs": [],
   "source": [
    "trainTokenizer = create_tokenizer(train)\n",
    "testTokenizer = create_tokenizer(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 365,
   "id": "beneficial-calvin",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "37606"
      ]
     },
     "execution_count": 365,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vocab_size = len(trainTokenizer.word_index) + 1\n",
    "vocab_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 368,
   "id": "terminal-sally",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "33\n"
     ]
    }
   ],
   "source": [
    "max_length1 = max([len(s.split()) for s in train.content])\n",
    "print(max_length1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 369,
   "id": "cardiovascular-education",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "34\n"
     ]
    }
   ],
   "source": [
    "max_length2 = max([len(s.split()) for s in test.content])\n",
    "print(max_length2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 370,
   "id": "recent-strip",
   "metadata": {},
   "outputs": [],
   "source": [
    "max_length =max_length2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 371,
   "id": "careful-easter",
   "metadata": {},
   "outputs": [],
   "source": [
    "Xtrain = encode_docs(tokenizer, max_length, train.content)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 372,
   "id": "micro-murray",
   "metadata": {},
   "outputs": [],
   "source": [
    "Xtest = encode_docs(tokenizer, max_length, test.content)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 374,
   "id": "younger-blast",
   "metadata": {},
   "outputs": [],
   "source": [
    "ytrain=train['target']\n",
    "ytest = test['target']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "frank-margin",
   "metadata": {},
   "outputs": [],
   "source": [
    "_, acc = model.evaluate(Xtrain, ytrain, verbose=0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dense-designer",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}