Hate Speech and Offensive Language Tweets.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dataset "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>Unnamed: 0</th>\n",
       "      <th>count</th>\n",
       "      <th>hate_speech</th>\n",
       "      <th>offensive_language</th>\n",
       "      <th>neither</th>\n",
       "      <th>class</th>\n",
       "      <th>tweet</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>2</td>\n",
       "      <td>!!! RT @mayasolovely: As a woman you shouldn't...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>!!!!! RT @mleew17: boy dats cold...tyga dwn ba...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>!!!!!!! RT @UrKindOfBrand Dawg!!!! RT @80sbaby...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>!!!!!!!!! RT @C_G_Anderson: @viva_based she lo...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>!!!!!!!!!!!!! RT @ShenikaRoberts: The shit you...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0  count  hate_speech  offensive_language  neither  class  \\\n",
       "0           0      3            0                   0        3      2   \n",
       "1           1      3            0                   3        0      1   \n",
       "2           2      3            0                   3        0      1   \n",
       "3           3      3            0                   2        1      1   \n",
       "4           4      6            0                   6        0      1   \n",
       "\n",
       "                                               tweet  \n",
       "0  !!! RT @mayasolovely: As a woman you shouldn't...  \n",
       "1  !!!!! RT @mleew17: boy dats cold...tyga dwn ba...  \n",
       "2  !!!!!!! RT @UrKindOfBrand Dawg!!!! RT @80sbaby...  \n",
       "3  !!!!!!!!! RT @C_G_Anderson: @viva_based she lo...  \n",
       "4  !!!!!!!!!!!!! RT @ShenikaRoberts: The shit you...  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv('data.csv')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "count = number of CrowdFlower users who coded each tweet (min is 3, sometimes more users coded a tweet when judgments were determined to be unreliable by CF).\n",
    "\n",
    "hate_speech = number of CF users who judged the tweet to be hate speech.\n",
    "\n",
    "offensive_language = number of CF users who judged the tweet to be offensive.\n",
    "\n",
    "neither = number of CF users who judged the tweet to be neither offensive nor non-offensive.\n",
    "\n",
    "class = class label for majority of CF users. 0 - hate speech 1 - offensive language 2 - neither"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "classes = ['Hate Speech','Offensive Language','None']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.drop(['count','hate_speech','offensive_language','neither','Unnamed: 0'],axis=1,inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>class</th>\n",
       "      <th>tweet</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2</td>\n",
       "      <td>!!! RT @mayasolovely: As a woman you shouldn't...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>!!!!! RT @mleew17: boy dats cold...tyga dwn ba...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>!!!!!!! RT @UrKindOfBrand Dawg!!!! RT @80sbaby...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>!!!!!!!!! RT @C_G_Anderson: @viva_based she lo...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>!!!!!!!!!!!!! RT @ShenikaRoberts: The shit you...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   class                                              tweet\n",
       "0      2  !!! RT @mayasolovely: As a woman you shouldn't...\n",
       "1      1  !!!!! RT @mleew17: boy dats cold...tyga dwn ba...\n",
       "2      1  !!!!!!! RT @UrKindOfBrand Dawg!!!! RT @80sbaby...\n",
       "3      1  !!!!!!!!! RT @C_G_Anderson: @viva_based she lo...\n",
       "4      1  !!!!!!!!!!!!! RT @ShenikaRoberts: The shit you..."
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(24783, 2)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hate Speech  :  1430\n",
      "Offensive Language  :  19190\n",
      "None  :  4163\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "labels = df['class']\n",
    "unique, counts = np.unique(labels, return_counts=True)\n",
    "values = list(zip(unique, counts))\n",
    "plt.bar(classes,counts)\n",
    "for i in values:\n",
    "    print(classes[i[0]],' : ',i[1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Extracting Tweets of all classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1430, 2)\n",
      "(19190, 2)\n",
      "(4163, 2)\n"
     ]
    }
   ],
   "source": [
    "hate_tweets = df[df['class']==0]\n",
    "offensive_tweets = df[df['class']==1]\n",
    "neither = df[df['class']==2]\n",
    "print(hate_tweets.shape)\n",
    "print(offensive_tweets.shape)\n",
    "print(neither.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Balancing the Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Taking around 12000 tweets for all the classes\n",
    "\n",
    "Repeating Hate tweets 8 times\n",
    "\n",
    "Repeating Neither tweets 2 times"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(11440, 2)\n",
      "(12000, 2)\n",
      "(12489, 2)\n"
     ]
    }
   ],
   "source": [
    "for i in range(3):\n",
    "    hate_tweets = pd.concat([hate_tweets,hate_tweets],ignore_index = True)\n",
    "neither = pd.concat([neither,neither,neither], ignore_index = True)\n",
    "offensive_tweets = offensive_tweets.iloc[0:12000,:]\n",
    "print(hate_tweets.shape)\n",
    "print(offensive_tweets.shape)\n",
    "print(neither.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(35929, 2)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.concat([hate_tweets,offensive_tweets,neither],ignore_index = True)\n",
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hate Speech  :  11440\n",
      "Offensive Language  :  12000\n",
      "None  :  12489\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "labels = df['class']\n",
    "unique, counts = np.unique(labels, return_counts=True)\n",
    "values = list(zip(unique, counts))\n",
    "plt.bar(classes,counts)\n",
    "for i in values:\n",
    "    print(classes[i[0]],' : ',i[1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>class</th>\n",
       "      <th>tweet</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@Blackman38Tide: @WhaleLookyHere @HowdyDowdy1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@CB_Baby24: @white_thunduh alsarabsss\" hes a ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@DevilGrimz: @VigxRArts you're fucking gay, b...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@MarkRoundtreeJr: LMFAOOOO I HATE BLACK PEOPL...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@NoChillPaz: \"At least I'm not a nigger\" http...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   class                                              tweet\n",
       "0      0  \"@Blackman38Tide: @WhaleLookyHere @HowdyDowdy1...\n",
       "1      0  \"@CB_Baby24: @white_thunduh alsarabsss\" hes a ...\n",
       "2      0  \"@DevilGrimz: @VigxRArts you're fucking gay, b...\n",
       "3      0  \"@MarkRoundtreeJr: LMFAOOOO I HATE BLACK PEOPL...\n",
       "4      0  \"@NoChillPaz: \"At least I'm not a nigger\" http..."
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tweets Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package wordnet to\n",
      "[nltk_data]     C:\\Users\\iampr\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package wordnet is already up-to-date!\n",
      "[nltk_data] Downloading package stopwords to\n",
      "[nltk_data]     C:\\Users\\iampr\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import nltk\n",
    "from nltk.stem import WordNetLemmatizer\n",
    "from nltk.corpus import stopwords\n",
    "import re\n",
    "nltk.download('wordnet')\n",
    "nltk.download('stopwords')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# dealing with Slangs\n",
    "d = {'luv':'love','wud':'would','lyk':'like','wateva':'whatever','ttyl':'talk to you later',\n",
    "               'kul':'cool','fyn':'fine','omg':'oh my god!','fam':'family','bruh':'brother',\n",
    "               'cud':'could','fud':'food', 'u': 'you', \n",
    "     'ur':'your', 'bday' : 'birthday', 'bihday' : 'birthday'} ## Need a huge dictionary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "stop_words = set(stopwords.words(\"english\"))\n",
    "stop_words.add('rt')\n",
    "stop_words.remove('not')\n",
    "lemmatizer = WordNetLemmatizer()\n",
    "giant_url_regex = ('http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|' '[!*\\(\\),]|(?:%[0-9a-fA-F][0-9a-fA-F]))+')\n",
    "mention_regex = '@[\\w\\-]+'\n",
    "\n",
    "def clean_text(text):\n",
    "    text = re.sub('\"', \"\", text)\n",
    "    text = re.sub(mention_regex, ' ',text) #removing all user names\n",
    "    text = re.sub(giant_url_regex, ' ', text)  #remocing the urls\n",
    "    text = text.lower()\n",
    "    text = re.sub(\"hm+\", \"\", text) #removing variants of hmmm\n",
    "    text = re.sub(\"[^a-z]+\", \" \", text) #removing all numbers, special chars like @,#,? etc\n",
    "    text = text.split()\n",
    "    text = [word for word in text if not word in stop_words]\n",
    "    text = [d[word] if word in d else word for word in text]  #replacing some slangs\n",
    "    text = [lemmatizer.lemmatize(token) for token in text]\n",
    "    text = [lemmatizer.lemmatize(token, \"v\") for token in text]\n",
    "    text = \" \".join(text)\n",
    "    return text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>class</th>\n",
       "      <th>tweet</th>\n",
       "      <th>processed_tweets</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@Blackman38Tide: @WhaleLookyHere @HowdyDowdy1...</td>\n",
       "      <td>queer gaywad</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@CB_Baby24: @white_thunduh alsarabsss\" hes a ...</td>\n",
       "      <td>alsarabsss he beaner smh tell he mexican</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@DevilGrimz: @VigxRArts you're fucking gay, b...</td>\n",
       "      <td>fuck gay blacklist hoe hold tehgodclan anyway</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@MarkRoundtreeJr: LMFAOOOO I HATE BLACK PEOPL...</td>\n",
       "      <td>lmfaoooo hate black people black people nigger</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>\"@NoChillPaz: \"At least I'm not a nigger\" http...</td>\n",
       "      <td>least not nigger lmfao</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   class                                              tweet  \\\n",
       "0      0  \"@Blackman38Tide: @WhaleLookyHere @HowdyDowdy1...   \n",
       "1      0  \"@CB_Baby24: @white_thunduh alsarabsss\" hes a ...   \n",
       "2      0  \"@DevilGrimz: @VigxRArts you're fucking gay, b...   \n",
       "3      0  \"@MarkRoundtreeJr: LMFAOOOO I HATE BLACK PEOPL...   \n",
       "4      0  \"@NoChillPaz: \"At least I'm not a nigger\" http...   \n",
       "\n",
       "                                 processed_tweets  \n",
       "0                                    queer gaywad  \n",
       "1        alsarabsss he beaner smh tell he mexican  \n",
       "2   fuck gay blacklist hoe hold tehgodclan anyway  \n",
       "3  lmfaoooo hate black people black people nigger  \n",
       "4                          least not nigger lmfao  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['processed_tweets'] = df.tweet.apply(lambda x: clean_text(x))   # df.review.map(clean_text) Also can be used\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(35929,)\n",
      "(35929,)\n"
     ]
    }
   ],
   "source": [
    "x = df.processed_tweets\n",
    "y = df['class']\n",
    "print(x.shape)\n",
    "print(y.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Word Frequency and Tweets Length"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The total words in the tweets are :  275540\n",
      "The total UNIQUE words in the tweets are :  14146\n"
     ]
    }
   ],
   "source": [
    "# finding unique words\n",
    "word_unique = []\n",
    "for i in x:\n",
    "    for j in i.split():\n",
    "        word_unique.append(j)\n",
    "unique, counts = np.unique(word_unique, return_counts=True)\n",
    "print(\"The total words in the tweets are : \", len(word_unique))\n",
    "print(\"The total UNIQUE words in the tweets are : \", len(unique))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The Average Length tweets are :  7.669013888502324\n",
      "The max length of tweets is :  28\n",
      "The min length of tweets is :  0\n"
     ]
    }
   ],
   "source": [
    "# finding length of tweets\n",
    "tweets_length = []\n",
    "for i in x:\n",
    "    tweets_length.append(len(i.split()))\n",
    "print(\"The Average Length tweets are : \",np.mean(tweets_length))\n",
    "print(\"The max length of tweets is : \", np.max(tweets_length))\n",
    "print(\"The min length of tweets is : \", np.min(tweets_length))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>35929.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>7.669014</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>3.989625</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>4.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>7.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>11.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>28.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  0\n",
       "count  35929.000000\n",
       "mean       7.669014\n",
       "std        3.989625\n",
       "min        0.000000\n",
       "25%        4.000000\n",
       "50%        7.000000\n",
       "75%       11.000000\n",
       "max       28.000000"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tweets_length = pd.DataFrame(tweets_length)\n",
    "tweets_length.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Top 20 Occuring Words with their frequency are:\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>bitch</td>\n",
       "      <td>9066</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>like</td>\n",
       "      <td>3817</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>get</td>\n",
       "      <td>3636</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hoe</td>\n",
       "      <td>3426</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>trash</td>\n",
       "      <td>3217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>fuck</td>\n",
       "      <td>3103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>nigga</td>\n",
       "      <td>2819</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>faggot</td>\n",
       "      <td>2239</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>as</td>\n",
       "      <td>2073</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>you</td>\n",
       "      <td>1851</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>pussy</td>\n",
       "      <td>1847</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>go</td>\n",
       "      <td>1773</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>not</td>\n",
       "      <td>1764</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>bird</td>\n",
       "      <td>1515</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>lol</td>\n",
       "      <td>1494</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>nigger</td>\n",
       "      <td>1459</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>say</td>\n",
       "      <td>1456</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>make</td>\n",
       "      <td>1373</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>amp</td>\n",
       "      <td>1329</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>white</td>\n",
       "      <td>1328</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         0     1\n",
       "0    bitch  9066\n",
       "1     like  3817\n",
       "2      get  3636\n",
       "3      hoe  3426\n",
       "4    trash  3217\n",
       "5     fuck  3103\n",
       "6    nigga  2819\n",
       "7   faggot  2239\n",
       "8       as  2073\n",
       "9      you  1851\n",
       "10   pussy  1847\n",
       "11      go  1773\n",
       "12     not  1764\n",
       "13    bird  1515\n",
       "14     lol  1494\n",
       "15  nigger  1459\n",
       "16     say  1456\n",
       "17    make  1373\n",
       "18     amp  1329\n",
       "19   white  1328"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Sorting the Unique words based on their Frequency\n",
    "col = list(zip(unique, counts))\n",
    "col = sorted(col, key = lambda x: x[1],reverse=True)\n",
    "col=pd.DataFrame(col)\n",
    "print(\"Top 20 Occuring Words with their frequency are:\")\n",
    "col.iloc[:20,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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>1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>14146.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>19.478298</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>128.082239</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>9.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>9066.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  1\n",
       "count  14146.000000\n",
       "mean      19.478298\n",
       "std      128.082239\n",
       "min        1.000000\n",
       "25%        1.000000\n",
       "50%        3.000000\n",
       "75%        9.000000\n",
       "max     9066.000000"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "col.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# TF-IDF Vectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8000\n",
      "(8000,)\n"
     ]
    }
   ],
   "source": [
    "vectorizer = TfidfVectorizer(max_features = 8000 )\n",
    "# tokenize and build vocab\n",
    "\n",
    "vectorizer.fit(x)\n",
    "# summarize\n",
    "\n",
    "print(len(vectorizer.vocabulary_))\n",
    "print(vectorizer.idf_.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(35929, 8000)\n"
     ]
    }
   ],
   "source": [
    "x_tfidf = vectorizer.transform(x).toarray()\n",
    "print(x_tfidf.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using ML with TF-IDF"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.svm import LinearSVC\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "from sklearn.model_selection import train_test_split as tts\n",
    "from sklearn.metrics import confusion_matrix, accuracy_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(34132, 8000)\n",
      "(1797, 8000)\n",
      "(34132,)\n",
      "(1797,)\n"
     ]
    }
   ],
   "source": [
    "x_train,x_test,y_train,y_test = tts(x_tfidf,y,test_size = 0.05)\n",
    "print(x_train.shape)\n",
    "print(x_test.shape)\n",
    "print(y_train.shape)\n",
    "print(y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SVM-1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\iampr\\AppData\\Roaming\\Python\\Python37\\site-packages\\sklearn\\svm\\_base.py:947: ConvergenceWarning: Liblinear failed to converge, increase the number of iterations.\n",
      "  \"the number of iterations.\", ConvergenceWarning)\n"
     ]
    }
   ],
   "source": [
    "svm_model = LinearSVC(class_weight='balanced',multi_class='crammer_singer',max_iter = -1).fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "svm_model_predict = svm_model.predict(x_test)\n",
    "svm_report = classification_report(y_test, svm_model_predict )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.96      0.92       562\n",
      "           1       0.95      0.85      0.89       604\n",
      "           2       0.96      0.98      0.97       631\n",
      "\n",
      "    accuracy                           0.93      1797\n",
      "   macro avg       0.93      0.93      0.93      1797\n",
      "weighted avg       0.93      0.93      0.93      1797\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(svm_report)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SVM-2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "svm_model_2 = LinearSVC(class_weight='balanced',C=1, penalty='l2', max_iter=1500,loss='squared_hinge',multi_class='ovr').fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "svm_model_predict_2 = svm_model_2.predict(x_test)\n",
    "svm_report_2 = classification_report(y_test, svm_model_predict_2 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.89      0.96      0.92       562\n",
      "           1       0.95      0.85      0.90       604\n",
      "           2       0.95      0.98      0.97       631\n",
      "\n",
      "    accuracy                           0.93      1797\n",
      "   macro avg       0.93      0.93      0.93      1797\n",
      "weighted avg       0.93      0.93      0.93      1797\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(svm_report_2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SGD Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "    svm_model_3 = SGDClassifier(n_jobs=-1,class_weight='balanced',penalty='l2').fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "svm_model_predict_3 = svm_model_3.predict(x_test)\n",
    "svm_report_3 = classification_report(y_test, svm_model_predict_3 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.81      0.81      0.81       562\n",
      "           1       0.87      0.81      0.84       604\n",
      "           2       0.90      0.95      0.92       631\n",
      "\n",
      "    accuracy                           0.86      1797\n",
      "   macro avg       0.86      0.86      0.86      1797\n",
      "weighted avg       0.86      0.86      0.86      1797\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(svm_report_3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed: 10.6min finished\n"
     ]
    }
   ],
   "source": [
    "logistic_reg_model = LogisticRegression(n_jobs = -1, penalty='l2', multi_class='multinomial',class_weight = 'balanced',verbose=1).fit(x_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr_model_predict = logistic_reg_model.predict(x_test)\n",
    "lr_model_report = classification_report(y_test, lr_model_predict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      0.93      0.90       562\n",
      "           1       0.92      0.84      0.88       604\n",
      "           2       0.95      0.97      0.96       631\n",
      "\n",
      "    accuracy                           0.91      1797\n",
      "   macro avg       0.91      0.91      0.91      1797\n",
      "weighted avg       0.92      0.91      0.91      1797\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(lr_model_report)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Padding and Tokenzation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "35929\n",
      "Mean is:  7.669013888502324\n",
      "Max is:  28\n",
      "Min is:  0\n"
     ]
    }
   ],
   "source": [
    "num_words = 8000\n",
    "embed_dim = 32\n",
    "tokenizer = Tokenizer(num_words=num_words,oov_token = \"<oov>\" )\n",
    "tokenizer.fit_on_texts(x)\n",
    "word_index=tokenizer.word_index\n",
    "sequences = tokenizer.texts_to_sequences(x)\n",
    "length=[]\n",
    "for i in sequences:\n",
    "    length.append(len(i))\n",
    "print(len(length))\n",
    "print(\"Mean is: \",np.mean(length))\n",
    "print(\"Max is: \",np.max(length))\n",
    "print(\"Min is: \",np.min(length))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(35929, 24)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pad_length = 24\n",
    "sequences = pad_sequences(sequences, maxlen = pad_length, truncating = 'pre', padding = 'post')\n",
    "sequences.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(34132, 24)\n",
      "(1797, 24)\n",
      "(34132,)\n",
      "(1797,)\n"
     ]
    }
   ],
   "source": [
    "x_train,x_test,y_train,y_test = tts(sequences,y,test_size = 0.05)\n",
    "print(x_train.shape)\n",
    "print(x_test.shape)\n",
    "print(y_train.shape)\n",
    "print(y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training Deep Learning Model on Embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import Dense, Embedding, Dropout , Activation, Flatten, SimpleRNN\n",
    "from keras.layers import GlobalMaxPool1D\n",
    "from keras.models import Model, Sequential\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"Twitter Hate Text Classification\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, 24, 32)            256000    \n",
      "_________________________________________________________________\n",
      "simple_rnn_1 (SimpleRNN)     (None, 24, 8)             328       \n",
      "_________________________________________________________________\n",
      "global_max_pooling1d_1 (Glob (None, 8)                 0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 20)                180       \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 20)                0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 3)                 63        \n",
      "=================================================================\n",
      "Total params: 256,571\n",
      "Trainable params: 256,571\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "recall = tf.keras.metrics.Recall()\n",
    "precision = tf.keras.metrics.Precision()\n",
    "\n",
    "model = Sequential([Embedding(num_words, embed_dim, input_length = pad_length),\n",
    "                   SimpleRNN(8, return_sequences = True),\n",
    "                   GlobalMaxPool1D(),\n",
    "                   Dense(20,activation = 'relu',kernel_initializer='he_uniform'),\n",
    "                   Dropout(0.25),\n",
    "                   Dense(3,activation = 'softmax')])\n",
    "model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "model.name = 'Twitter Hate Text Classification'\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\iampr\\anaconda3\\envs\\tensorflow\\lib\\site-packages\\tensorflow_core\\python\\framework\\indexed_slices.py:433: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory.\n",
      "  \"Converting sparse IndexedSlices to a dense Tensor of unknown shape. \"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 32425 samples, validate on 1707 samples\n",
      "Epoch 1/5\n",
      "32425/32425 [==============================] - 69s 2ms/step - loss: 0.5078 - accuracy: 0.8026 - val_loss: 0.2481 - val_accuracy: 0.9162\n",
      "Epoch 2/5\n",
      "32425/32425 [==============================] - 56s 2ms/step - loss: 0.1920 - accuracy: 0.9424 - val_loss: 0.1875 - val_accuracy: 0.9361\n",
      "Epoch 3/5\n",
      "32425/32425 [==============================] - 61s 2ms/step - loss: 0.1198 - accuracy: 0.9660 - val_loss: 0.1486 - val_accuracy: 0.9549\n",
      "Epoch 4/5\n",
      "32425/32425 [==============================] - 66s 2ms/step - loss: 0.0833 - accuracy: 0.9779 - val_loss: 0.1587 - val_accuracy: 0.9596\n",
      "Epoch 5/5\n",
      "32425/32425 [==============================] - 57s 2ms/step - loss: 0.0632 - accuracy: 0.9834 - val_loss: 0.1509 - val_accuracy: 0.9572\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(x = x_train, y = y_train, epochs = 5,validation_split = 0.05)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_acc = history.history['accuracy']\n",
    "valid_acc = history.history['val_accuracy']\n",
    "train_loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib\n",
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x2221c43eec8>"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig=plt.figure(figsize=(10,10))\n",
    "fig.add_subplot(2, 1, 1)\n",
    "plt.title('Loss')\n",
    "plt.grid()\n",
    "plt.plot(train_loss, color='blue', label='Train')\n",
    "plt.plot(val_loss, color='red', label='Validation')\n",
    "plt.legend()\n",
    "# plot accuracy\n",
    "fig.add_subplot(2, 1, 2)\n",
    "plt.title('Classification Accuracy')\n",
    "plt.grid()\n",
    "plt.plot(train_acc, color='blue', label='Train')\n",
    "plt.plot(valid_acc, color='red', label='Validation')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1797/1797 [==============================] - 0s 262us/step\n"
     ]
    }
   ],
   "source": [
    "evaluate = model.evaluate(x_test,y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Acuracy is : 95.99 %\n",
      "Test Loss is : 0.1506\n"
     ]
    }
   ],
   "source": [
    "print(\"Test Acuracy is : {:.2f} %\".format(evaluate[1]*100))\n",
    "print(\"Test Loss is : {:.4f}\".format(evaluate[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = model.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "predict = []\n",
    "for i in predictions:\n",
    "    predict.append(np.argmax(i))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "cm = confusion_matrix(predict,y_test)\n",
    "acc = accuracy_score(predict,y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The Confusion matrix is: \n",
      " [[564  35   0]\n",
      " [  7 529   7]\n",
      " [  1  22 632]]\n"
     ]
    }
   ],
   "source": [
    "print(\"The Confusion matrix is: \\n\",cm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "95.99332220367279\n"
     ]
    }
   ],
   "source": [
    "print(acc*100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.94      0.99      0.96       572\n",
      "           1       0.97      0.90      0.94       586\n",
      "           2       0.96      0.99      0.98       639\n",
      "\n",
      "    accuracy                           0.96      1797\n",
      "   macro avg       0.96      0.96      0.96      1797\n",
      "weighted avg       0.96      0.96      0.96      1797\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(metrics.classification_report(y_test, predict))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tensorflow",
   "language": "python",
   "name": "tensorflow"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}