tweet_sentiment_classifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "MP8guU7geWYg"
   },
   "source": [
    "# Libraries & Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:47:42.200100Z",
     "start_time": "2024-09-23T11:47:41.760235Z"
    }
   },
   "outputs": [],
   "source": [
    "# !pip install numpy==1.23.5\n",
    "# !pip install scikit-learn\n",
    "# !pip install pandas\n",
    "# !pip install matplotlib\n",
    "# !pip install nltk\n",
    "# !pip install emoji\n",
    "# !pip install -U keras-tuner\n",
    "# !pip install tensorflow\n",
    "# !pip install kaggle\n",
    "# !pip install -U spacy\n",
    "# !pip install gensim\n",
    "#\n",
    "# !rm -r KT_dir_2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:32:03.168441Z",
     "start_time": "2024-09-23T13:32:03.132134Z"
    }
   },
   "outputs": [],
   "source": [
    "import nltk\n",
    "import tensorflow as tf\n",
    "import keras_tuner as kt\n",
    "import spacy\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os, json, re, time, emoji, kaggle, string, locale\n",
    "from tqdm import tqdm\n",
    "import gensim.downloader as api\n",
    "\n",
    "\n",
    "from spacy.lang.en.stop_words import STOP_WORDS\n",
    "\n",
    "from nltk.tag.perceptron import PerceptronTagger\n",
    "from nltk.tokenize import TweetTokenizer\n",
    "from nltk.stem import WordNetLemmatizer\n",
    "from nltk.corpus import wordnet, stopwords\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import (\n",
    "    precision_recall_curve,\n",
    "    auc,\n",
    "    accuracy_score,\n",
    "    ConfusionMatrixDisplay,\n",
    "    f1_score,\n",
    "    recall_score,\n",
    "    precision_score,\n",
    "    classification_report,\n",
    ")\n",
    "from sklearn.dummy import DummyClassifier\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.linear_model import LogisticRegressionCV\n",
    "\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Dropout, Input, Layer, Embedding, Bidirectional, GRU\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.metrics import CategoricalAccuracy\n",
    "from tensorflow.keras.models import load_model\n",
    "from tensorflow.keras.callbacks import EarlyStopping, Callback\n",
    "from tensorflow.keras.layers import TextVectorization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T19:45:35.009188Z",
     "start_time": "2024-09-23T19:45:34.996824Z"
    }
   },
   "outputs": [],
   "source": [
    "# os.environ[\"TF_CPP_MIN_LOG_LEVEL\"] = \"3\"\n",
    "#\n",
    "# nltk.download(\"punkt\")\n",
    "# nltk.download(\"stopwords\")\n",
    "# nltk.download(\"wordnet\")\n",
    "# nltk.download(\"omw-1.4\")\n",
    "# nltk.download(\"averaged_perceptron_tagger\")\n",
    "# nltk.download('averaged_perceptron_tagger_eng')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "outputs": [],
   "source": [
    "# !python -m spacy download en_core_web_sm"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:40.215198Z",
     "start_time": "2024-09-23T11:48:40.176711Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "id": "GybjyfVfeskj",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:41.846931Z",
     "start_time": "2024-09-23T11:48:40.512590Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Warning: Looks like you're using an outdated API Version, please consider updating (server 1.6.17 / client 1.6.14)\r\n",
      "Dataset URL: https://www.kaggle.com/datasets/jp797498e/twitter-entity-sentiment-analysis\r\n",
      "License(s): CC0-1.0\r\n",
      "twitter-entity-sentiment-analysis.zip: Skipping, found more recently modified local copy (use --force to force download)\r\n"
     ]
    }
   ],
   "source": [
    "# Ensure the Kaggle directory exists\n",
    "os.makedirs(os.path.expanduser(\"~/.kaggle\"), exist_ok=True)\n",
    "\n",
    "# Copy kaggle.json to the right location if it is not already there\n",
    "# Uncomment and modify the below lines if you need to place your kaggle.json file in the right directory\n",
    "# import shutil\n",
    "# shutil.copy('path_to_your_kaggle.json', os.path.expanduser('~/.kaggle/kaggle.json'))\n",
    "\n",
    "# Download the dataset using Kaggle API\n",
    "!kaggle datasets download -d jp797498e/twitter-entity-sentiment-analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:42.232436Z",
     "start_time": "2024-09-23T11:48:41.848715Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mkdir: /Users/antonal/.kaggle: File exists\r\n"
     ]
    }
   ],
   "source": [
    "!mkdir ~/.kaggle\n",
    "!touch ~/.kaggle/kaggle.json\n",
    "\n",
    "api_token = {\"username\": \"{username}\", \"key\": \"{key}\"}\n",
    "\n",
    "with open(\"/Users/antonal/.kaggle/kaggle.json\", \"w\") as file:\n",
    "    json.dump(api_token, file)\n",
    "\n",
    "!chmod 600 ~/.kaggle/kaggle.json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T19:45:48.425805Z",
     "start_time": "2024-09-23T19:45:48.405664Z"
    }
   },
   "outputs": [],
   "source": [
    "kaggle.api.authenticate()\n",
    "kaggle.api.competition_download_files(\"twitter-sentiment-analysis2\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T19:45:55.935758Z",
     "start_time": "2024-09-23T19:45:55.905139Z"
    }
   },
   "outputs": [],
   "source": [
    "!unzip twitter-sentiment-analysis2.zip"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "### Initialise dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "id": "lTf7s7zSetW4",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:44.487015Z",
     "start_time": "2024-09-23T11:48:44.385418Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "      Sentiment                                      SentimentText\n0             1  #mygarage - new hashtag for the interesting ri...\n1             0  #Haveyouever why you and your mom just can't g...\n2             0                                #BGT That poor kid \n3             1  #haveyouever had cream cheese &amp; french fri...\n4             1       =- trying to figure out this Twitter thingie\n...         ...                                                ...\n9995          0   not even one single reply.what are people loo...\n9996          1  #phish has an awesome fog machine. oh wait.  n...\n9997          1   gotta love her. glad she had fun. looking for...\n9998          1  &quot;I'm wore out like the waist band on some...\n9999          1  &quot;how do you like them apples&quot; is fro...\n\n[10000 rows x 2 columns]",
      "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>Sentiment</th>\n      <th>SentimentText</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>#mygarage - new hashtag for the interesting ri...</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0</td>\n      <td>#Haveyouever why you and your mom just can't g...</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0</td>\n      <td>#BGT That poor kid</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>#haveyouever had cream cheese &amp;amp; french fri...</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1</td>\n      <td>=- trying to figure out this Twitter thingie</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>9995</th>\n      <td>0</td>\n      <td>not even one single reply.what are people loo...</td>\n    </tr>\n    <tr>\n      <th>9996</th>\n      <td>1</td>\n      <td>#phish has an awesome fog machine. oh wait.  n...</td>\n    </tr>\n    <tr>\n      <th>9997</th>\n      <td>1</td>\n      <td>gotta love her. glad she had fun. looking for...</td>\n    </tr>\n    <tr>\n      <th>9998</th>\n      <td>1</td>\n      <td>&amp;quot;I'm wore out like the waist band on some...</td>\n    </tr>\n    <tr>\n      <th>9999</th>\n      <td>1</td>\n      <td>&amp;quot;how do you like them apples&amp;quot; is fro...</td>\n    </tr>\n  </tbody>\n</table>\n<p>10000 rows × 2 columns</p>\n</div>"
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Read the CSV file 'train.csv', selecting only the 'Sentiment' and 'SentimentText' columns.\n",
    "df_tweet_sentiment = pd.read_csv(\n",
    "    \"./train.csv\",\n",
    "    encoding=\"ISO-8859-1\",\n",
    "    usecols=[\"Sentiment\", \"SentimentText\"],\n",
    "    index_col=False,\n",
    ")\n",
    "\n",
    "# Filtering the DataFrame to get the first 5000 rows where the sentiment is 0 (negative)\n",
    "df_negative_tweets = df_tweet_sentiment[df_tweet_sentiment[\"Sentiment\"] == 0].head(5000)\n",
    "\n",
    "# Filtering the DataFrame to get the first 5000 rows where the sentiment is 1 (positive)\n",
    "df_positive_tweets = df_tweet_sentiment[df_tweet_sentiment[\"Sentiment\"] == 1].head(5000)\n",
    "\n",
    "# Concatenate the two filtered DataFrames into a single DataFrame\n",
    "df_tweet_sentiment = pd.concat([df_negative_tweets, df_positive_tweets])\n",
    "\n",
    "# Shuffle the concatenated DataFrame randomly and reset the index\n",
    "df_tweet_sentiment = df_tweet_sentiment.sample(frac=1).reset_index(drop=True)\n",
    "df_tweet_sentiment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "pm6_60nkfACQ",
    "outputId": "7c408f39-028d-4043-ded4-270497fd7369",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:46.093993Z",
     "start_time": "2024-09-23T11:48:46.012950Z"
    }
   },
   "outputs": [],
   "source": [
    "# Initialize a WordNetLemmatizer, which is used to reduce words to their base or root form, e.g., (churches -> church, abaci -> abacus)\n",
    "lemmatizer = WordNetLemmatizer()\n",
    "\n",
    "# Initialize a PerceptronTagger, which is a part-of-speech tagger used to label words with their corresponding parts of speech\n",
    "tagger = PerceptronTagger()\n",
    "\n",
    "# Initialize a TweetTokenizer, which is specifically designed to handle tokenizing tweets (short text with special characters and emoticons)\n",
    "tweet_wt = TweetTokenizer()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "wp_eqjYHeejp"
   },
   "source": [
    "# Preprocess dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "id": "CgO7FZs4eUk6",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:49.459330Z",
     "start_time": "2024-09-23T11:48:49.434099Z"
    }
   },
   "outputs": [],
   "source": [
    "# Lemmatization helper function\n",
    "def nltk_pos_tagger(nltk_tag):\n",
    "    if nltk_tag.startswith(\"J\"):\n",
    "        return wordnet.ADJ\n",
    "    elif nltk_tag.startswith(\"V\"):\n",
    "        return wordnet.VERB\n",
    "    elif nltk_tag.startswith(\"N\"):\n",
    "        return wordnet.NOUN\n",
    "    elif nltk_tag.startswith(\"R\"):\n",
    "        return wordnet.ADV\n",
    "    else:\n",
    "        return None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "id": "Ow1-4g8MekuY",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:49.906099Z",
     "start_time": "2024-09-23T11:48:49.879753Z"
    }
   },
   "outputs": [],
   "source": [
    "def replace_special_tokens(sequence):\n",
    "    # Replace URLs with the token \"<url>\"\n",
    "    sequence = re.sub(\n",
    "        \"\"\"http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|[!*\\(\\),]|(?:%[0-9a-fA-F]\n",
    "        [0-9a-fA-F]))+\"\"\",\n",
    "        \"<url>\",\n",
    "        sequence,\n",
    "    )\n",
    "\n",
    "    # Replace mentions (words starting with @) with the token \"<user>\"\n",
    "    sequence = re.sub(\"@\\S+\", \"<user>\", sequence)\n",
    "\n",
    "    # Replace hashtags (words starting with #) with the token \"<hashtag>\"\n",
    "    sequence = re.sub(\"#\\S+\", \"<hashtag>\", sequence)\n",
    "\n",
    "    # Replace numbers (including decimals) with the token \"<number>\"\n",
    "    sequence = re.sub(\"([+-]?([0-9]+([.][0-9]*)?|[.][0-9]+))\", \"<number>\", sequence)\n",
    "\n",
    "    # Convert emojis to text representation (e.g., 😊 to :smiling_face_with_smiling_eyes:)\n",
    "    sequence = emoji.demojize(sequence)  # remove emojis\n",
    "\n",
    "    # Remove emoticons (represented as :word:) from the sequence\n",
    "    sequence = re.sub(r\"\\s*:[\\w]+\", \" \", sequence)  # remove emoticons\n",
    "\n",
    "    # Convert the entire sequence to lowercase\n",
    "    sequence = sequence.lower()\n",
    "\n",
    "    return sequence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "id": "wczDPnTXenLE",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:53.596006Z",
     "start_time": "2024-09-23T11:48:50.354954Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "      Sentiment                                      SentimentText  \\\n0             1  #mygarage - new hashtag for the interesting ri...   \n1             0  #Haveyouever why you and your mom just can't g...   \n2             0                                #BGT That poor kid    \n3             1  #haveyouever had cream cheese &amp; french fri...   \n4             1       =- trying to figure out this Twitter thingie   \n...         ...                                                ...   \n9995          0   not even one single reply.what are people loo...   \n9996          1  #phish has an awesome fog machine. oh wait.  n...   \n9997          1   gotta love her. glad she had fun. looking for...   \n9998          1  &quot;I'm wore out like the waist band on some...   \n9999          1  &quot;how do you like them apples&quot; is fro...   \n\n                                         processed_text  \\\n0     <hashtag> - new hashtag for the interesting ri...   \n1     <hashtag> why you and your mom just can't get ...   \n2                              <hashtag> that poor kid    \n3     <hashtag> had cream cheese &amp; french fries ...   \n4          =- trying to figure out this twitter thingie   \n...                                                 ...   \n9995   not even one single reply.what are people loo...   \n9996  <hashtag> has an awesome fog machine. oh wait....   \n9997   gotta love her. glad she had fun. looking for...   \n9998  &quot;i'm wore out like the waist band on some...   \n9999  &quot;how do you like them apples&quot; is fro...   \n\n                                         tokenized_text  \\\n0     [<hashtag>, -, new, hashtag, for, the, interes...   \n1     [<hashtag>, why, you, and, your, mom, just, ca...   \n2                          [<hashtag>, that, poor, kid]   \n3     [<hashtag>, had, cream, cheese, &, french, fri...   \n4     [=, -, trying, to, figure, out, this, twitter,...   \n...                                                 ...   \n9995  [not, even, one, single, reply.what, are, peop...   \n9996  [<hashtag>, has, an, awesome, fog, machine, .,...   \n9997  [gotta, love, her, ., glad, she, had, fun, ., ...   \n9998  [\", i'm, wore, out, like, the, waist, band, on...   \n9999  [\", how, do, you, like, them, apples, \", is, f...   \n\n                                              text_tags  \n0     [(<hashtag>, SYM), (-, :), (new, JJ), (hashtag...  \n1     [(<hashtag>, JJ), (why, WRB), (you, PRP), (and...  \n2     [(<hashtag>, NN), (that, WDT), (poor, JJ), (ki...  \n3     [(<hashtag>, NN), (had, VBD), (cream, VBN), (c...  \n4     [(=, SYM), (-, :), (trying, VBG), (to, TO), (f...  \n...                                                 ...  \n9995  [(not, RB), (even, RB), (one, CD), (single, JJ...  \n9996  [(<hashtag>, NN), (has, VBZ), (an, DT), (aweso...  \n9997  [(gotta, NN), (love, VB), (her, PRP$), (., .),...  \n9998  [(\", JJ), (i'm, NN), (wore, VBD), (out, RP), (...  \n9999  [(\", VB), (how, WRB), (do, VB), (you, PRP), (l...  \n\n[10000 rows x 5 columns]",
      "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>Sentiment</th>\n      <th>SentimentText</th>\n      <th>processed_text</th>\n      <th>tokenized_text</th>\n      <th>text_tags</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>#mygarage - new hashtag for the interesting ri...</td>\n      <td>&lt;hashtag&gt; - new hashtag for the interesting ri...</td>\n      <td>[&lt;hashtag&gt;, -, new, hashtag, for, the, interes...</td>\n      <td>[(&lt;hashtag&gt;, SYM), (-, :), (new, JJ), (hashtag...</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0</td>\n      <td>#Haveyouever why you and your mom just can't g...</td>\n      <td>&lt;hashtag&gt; why you and your mom just can't get ...</td>\n      <td>[&lt;hashtag&gt;, why, you, and, your, mom, just, ca...</td>\n      <td>[(&lt;hashtag&gt;, JJ), (why, WRB), (you, PRP), (and...</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0</td>\n      <td>#BGT That poor kid</td>\n      <td>&lt;hashtag&gt; that poor kid</td>\n      <td>[&lt;hashtag&gt;, that, poor, kid]</td>\n      <td>[(&lt;hashtag&gt;, NN), (that, WDT), (poor, JJ), (ki...</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>#haveyouever had cream cheese &amp;amp; french fri...</td>\n      <td>&lt;hashtag&gt; had cream cheese &amp;amp; french fries ...</td>\n      <td>[&lt;hashtag&gt;, had, cream, cheese, &amp;, french, fri...</td>\n      <td>[(&lt;hashtag&gt;, NN), (had, VBD), (cream, VBN), (c...</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1</td>\n      <td>=- trying to figure out this Twitter thingie</td>\n      <td>=- trying to figure out this twitter thingie</td>\n      <td>[=, -, trying, to, figure, out, this, twitter,...</td>\n      <td>[(=, SYM), (-, :), (trying, VBG), (to, TO), (f...</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>9995</th>\n      <td>0</td>\n      <td>not even one single reply.what are people loo...</td>\n      <td>not even one single reply.what are people loo...</td>\n      <td>[not, even, one, single, reply.what, are, peop...</td>\n      <td>[(not, RB), (even, RB), (one, CD), (single, JJ...</td>\n    </tr>\n    <tr>\n      <th>9996</th>\n      <td>1</td>\n      <td>#phish has an awesome fog machine. oh wait.  n...</td>\n      <td>&lt;hashtag&gt; has an awesome fog machine. oh wait....</td>\n      <td>[&lt;hashtag&gt;, has, an, awesome, fog, machine, .,...</td>\n      <td>[(&lt;hashtag&gt;, NN), (has, VBZ), (an, DT), (aweso...</td>\n    </tr>\n    <tr>\n      <th>9997</th>\n      <td>1</td>\n      <td>gotta love her. glad she had fun. looking for...</td>\n      <td>gotta love her. glad she had fun. looking for...</td>\n      <td>[gotta, love, her, ., glad, she, had, fun, ., ...</td>\n      <td>[(gotta, NN), (love, VB), (her, PRP$), (., .),...</td>\n    </tr>\n    <tr>\n      <th>9998</th>\n      <td>1</td>\n      <td>&amp;quot;I'm wore out like the waist band on some...</td>\n      <td>&amp;quot;i'm wore out like the waist band on some...</td>\n      <td>[\", i'm, wore, out, like, the, waist, band, on...</td>\n      <td>[(\", JJ), (i'm, NN), (wore, VBD), (out, RP), (...</td>\n    </tr>\n    <tr>\n      <th>9999</th>\n      <td>1</td>\n      <td>&amp;quot;how do you like them apples&amp;quot; is fro...</td>\n      <td>&amp;quot;how do you like them apples&amp;quot; is fro...</td>\n      <td>[\", how, do, you, like, them, apples, \", is, f...</td>\n      <td>[(\", VB), (how, WRB), (do, VB), (you, PRP), (l...</td>\n    </tr>\n  </tbody>\n</table>\n<p>10000 rows × 5 columns</p>\n</div>"
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Apply the replace_special_tokens function to the 'SentimentText' column to preprocess the text\n",
    "df_tweet_sentiment[\"processed_text\"] = df_tweet_sentiment[\"SentimentText\"].apply(\n",
    "    replace_special_tokens\n",
    ")\n",
    "\n",
    "# Tokenize the processed text using the TweetTokenizer\n",
    "df_tweet_sentiment[\"tokenized_text\"] = df_tweet_sentiment[\"processed_text\"].apply(\n",
    "    lambda x: tweet_wt.tokenize(x)\n",
    ")\n",
    "\n",
    "# Tag the tokenized text with part-of-speech tags using the PerceptronTagger\n",
    "df_tweet_sentiment[\"text_tags\"] = df_tweet_sentiment[\"tokenized_text\"].apply(\n",
    "    lambda x: tagger.tag(x)\n",
    ")\n",
    "df_tweet_sentiment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "id": "epLzK6k9f56M",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:53.596077Z",
     "start_time": "2024-09-23T11:48:53.593314Z"
    }
   },
   "outputs": [],
   "source": [
    "# Function to lemmatize a sequence of tokens with POS tags\n",
    "def lemma(sequence):\n",
    "    return [\n",
    "        lemmatizer.lemmatize(token, pos=nltk_pos_tagger(pos_tag))\n",
    "        if nltk_pos_tagger(pos_tag) != None\n",
    "        else lemmatizer.lemmatize(token)\n",
    "        for token, pos_tag in sequence\n",
    "    ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:56.214047Z",
     "start_time": "2024-09-23T11:48:55.975641Z"
    }
   },
   "outputs": [],
   "source": [
    "# Apply the lemma function to the 'text_tags' column of the dataframe\n",
    "# 'text_tags' is assumed to be a list of (token, pos_tag) tuples for each tweet\n",
    "df_tweet_sentiment[\"lemmatized_text\"] = df_tweet_sentiment[\"text_tags\"].apply(\n",
    "    lambda x: lemma(x)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:48:56.800448Z",
     "start_time": "2024-09-23T11:48:56.775388Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "      Sentiment                                      SentimentText  \\\n0             1  #mygarage - new hashtag for the interesting ri...   \n1             0  #Haveyouever why you and your mom just can't g...   \n2             0                                #BGT That poor kid    \n3             1  #haveyouever had cream cheese &amp; french fri...   \n4             1       =- trying to figure out this Twitter thingie   \n...         ...                                                ...   \n9995          0   not even one single reply.what are people loo...   \n9996          1  #phish has an awesome fog machine. oh wait.  n...   \n9997          1   gotta love her. glad she had fun. looking for...   \n9998          1  &quot;I'm wore out like the waist band on some...   \n9999          1  &quot;how do you like them apples&quot; is fro...   \n\n                                         processed_text  \\\n0     <hashtag> - new hashtag for the interesting ri...   \n1     <hashtag> why you and your mom just can't get ...   \n2                              <hashtag> that poor kid    \n3     <hashtag> had cream cheese &amp; french fries ...   \n4          =- trying to figure out this twitter thingie   \n...                                                 ...   \n9995   not even one single reply.what are people loo...   \n9996  <hashtag> has an awesome fog machine. oh wait....   \n9997   gotta love her. glad she had fun. looking for...   \n9998  &quot;i'm wore out like the waist band on some...   \n9999  &quot;how do you like them apples&quot; is fro...   \n\n                                         tokenized_text  \\\n0     [<hashtag>, -, new, hashtag, for, the, interes...   \n1     [<hashtag>, why, you, and, your, mom, just, ca...   \n2                          [<hashtag>, that, poor, kid]   \n3     [<hashtag>, had, cream, cheese, &, french, fri...   \n4     [=, -, trying, to, figure, out, this, twitter,...   \n...                                                 ...   \n9995  [not, even, one, single, reply.what, are, peop...   \n9996  [<hashtag>, has, an, awesome, fog, machine, .,...   \n9997  [gotta, love, her, ., glad, she, had, fun, ., ...   \n9998  [\", i'm, wore, out, like, the, waist, band, on...   \n9999  [\", how, do, you, like, them, apples, \", is, f...   \n\n                                              text_tags  \\\n0     [(<hashtag>, SYM), (-, :), (new, JJ), (hashtag...   \n1     [(<hashtag>, JJ), (why, WRB), (you, PRP), (and...   \n2     [(<hashtag>, NN), (that, WDT), (poor, JJ), (ki...   \n3     [(<hashtag>, NN), (had, VBD), (cream, VBN), (c...   \n4     [(=, SYM), (-, :), (trying, VBG), (to, TO), (f...   \n...                                                 ...   \n9995  [(not, RB), (even, RB), (one, CD), (single, JJ...   \n9996  [(<hashtag>, NN), (has, VBZ), (an, DT), (aweso...   \n9997  [(gotta, NN), (love, VB), (her, PRP$), (., .),...   \n9998  [(\", JJ), (i'm, NN), (wore, VBD), (out, RP), (...   \n9999  [(\", VB), (how, WRB), (do, VB), (you, PRP), (l...   \n\n                                        lemmatized_text  \\\n0     [<hashtag>, -, new, hashtag, for, the, interes...   \n1     [<hashtag>, why, you, and, your, mom, just, ca...   \n2                          [<hashtag>, that, poor, kid]   \n3     [<hashtag>, have, cream, cheese, &, french, fr...   \n4     [=, -, try, to, figure, out, this, twitter, th...   \n...                                                 ...   \n9995  [not, even, one, single, reply.what, be, peopl...   \n9996  [<hashtag>, have, an, awesome, fog, machine, ....   \n9997  [gotta, love, her, ., glad, she, have, fun, .,...   \n9998  [\", i'm, wear, out, like, the, waist, band, on...   \n9999  [\", how, do, you, like, them, apple, \", be, fr...   \n\n                                 lemmatized_text_joined  \n0     <hashtag> - new hashtag for the interesting ri...  \n1     <hashtag> why you and your mom just can't get ...  \n2                               <hashtag> that poor kid  \n3     <hashtag> have cream cheese & french fry ( sen...  \n4            = - try to figure out this twitter thingie  \n...                                                 ...  \n9995  not even one single reply.what be people look ...  \n9996  <hashtag> have an awesome fog machine . oh wai...  \n9997  gotta love her . glad she have fun . look forw...  \n9998  \" i'm wear out like the waist band on some old...  \n9999  \" how do you like them apple \" be from the mov...  \n\n[10000 rows x 7 columns]",
      "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>Sentiment</th>\n      <th>SentimentText</th>\n      <th>processed_text</th>\n      <th>tokenized_text</th>\n      <th>text_tags</th>\n      <th>lemmatized_text</th>\n      <th>lemmatized_text_joined</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>#mygarage - new hashtag for the interesting ri...</td>\n      <td>&lt;hashtag&gt; - new hashtag for the interesting ri...</td>\n      <td>[&lt;hashtag&gt;, -, new, hashtag, for, the, interes...</td>\n      <td>[(&lt;hashtag&gt;, SYM), (-, :), (new, JJ), (hashtag...</td>\n      <td>[&lt;hashtag&gt;, -, new, hashtag, for, the, interes...</td>\n      <td>&lt;hashtag&gt; - new hashtag for the interesting ri...</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0</td>\n      <td>#Haveyouever why you and your mom just can't g...</td>\n      <td>&lt;hashtag&gt; why you and your mom just can't get ...</td>\n      <td>[&lt;hashtag&gt;, why, you, and, your, mom, just, ca...</td>\n      <td>[(&lt;hashtag&gt;, JJ), (why, WRB), (you, PRP), (and...</td>\n      <td>[&lt;hashtag&gt;, why, you, and, your, mom, just, ca...</td>\n      <td>&lt;hashtag&gt; why you and your mom just can't get ...</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0</td>\n      <td>#BGT That poor kid</td>\n      <td>&lt;hashtag&gt; that poor kid</td>\n      <td>[&lt;hashtag&gt;, that, poor, kid]</td>\n      <td>[(&lt;hashtag&gt;, NN), (that, WDT), (poor, JJ), (ki...</td>\n      <td>[&lt;hashtag&gt;, that, poor, kid]</td>\n      <td>&lt;hashtag&gt; that poor kid</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>#haveyouever had cream cheese &amp;amp; french fri...</td>\n      <td>&lt;hashtag&gt; had cream cheese &amp;amp; french fries ...</td>\n      <td>[&lt;hashtag&gt;, had, cream, cheese, &amp;, french, fri...</td>\n      <td>[(&lt;hashtag&gt;, NN), (had, VBD), (cream, VBN), (c...</td>\n      <td>[&lt;hashtag&gt;, have, cream, cheese, &amp;, french, fr...</td>\n      <td>&lt;hashtag&gt; have cream cheese &amp; french fry ( sen...</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1</td>\n      <td>=- trying to figure out this Twitter thingie</td>\n      <td>=- trying to figure out this twitter thingie</td>\n      <td>[=, -, trying, to, figure, out, this, twitter,...</td>\n      <td>[(=, SYM), (-, :), (trying, VBG), (to, TO), (f...</td>\n      <td>[=, -, try, to, figure, out, this, twitter, th...</td>\n      <td>= - try to figure out this twitter thingie</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>9995</th>\n      <td>0</td>\n      <td>not even one single reply.what are people loo...</td>\n      <td>not even one single reply.what are people loo...</td>\n      <td>[not, even, one, single, reply.what, are, peop...</td>\n      <td>[(not, RB), (even, RB), (one, CD), (single, JJ...</td>\n      <td>[not, even, one, single, reply.what, be, peopl...</td>\n      <td>not even one single reply.what be people look ...</td>\n    </tr>\n    <tr>\n      <th>9996</th>\n      <td>1</td>\n      <td>#phish has an awesome fog machine. oh wait.  n...</td>\n      <td>&lt;hashtag&gt; has an awesome fog machine. oh wait....</td>\n      <td>[&lt;hashtag&gt;, has, an, awesome, fog, machine, .,...</td>\n      <td>[(&lt;hashtag&gt;, NN), (has, VBZ), (an, DT), (aweso...</td>\n      <td>[&lt;hashtag&gt;, have, an, awesome, fog, machine, ....</td>\n      <td>&lt;hashtag&gt; have an awesome fog machine . oh wai...</td>\n    </tr>\n    <tr>\n      <th>9997</th>\n      <td>1</td>\n      <td>gotta love her. glad she had fun. looking for...</td>\n      <td>gotta love her. glad she had fun. looking for...</td>\n      <td>[gotta, love, her, ., glad, she, had, fun, ., ...</td>\n      <td>[(gotta, NN), (love, VB), (her, PRP$), (., .),...</td>\n      <td>[gotta, love, her, ., glad, she, have, fun, .,...</td>\n      <td>gotta love her . glad she have fun . look forw...</td>\n    </tr>\n    <tr>\n      <th>9998</th>\n      <td>1</td>\n      <td>&amp;quot;I'm wore out like the waist band on some...</td>\n      <td>&amp;quot;i'm wore out like the waist band on some...</td>\n      <td>[\", i'm, wore, out, like, the, waist, band, on...</td>\n      <td>[(\", JJ), (i'm, NN), (wore, VBD), (out, RP), (...</td>\n      <td>[\", i'm, wear, out, like, the, waist, band, on...</td>\n      <td>\" i'm wear out like the waist band on some old...</td>\n    </tr>\n    <tr>\n      <th>9999</th>\n      <td>1</td>\n      <td>&amp;quot;how do you like them apples&amp;quot; is fro...</td>\n      <td>&amp;quot;how do you like them apples&amp;quot; is fro...</td>\n      <td>[\", how, do, you, like, them, apples, \", is, f...</td>\n      <td>[(\", VB), (how, WRB), (do, VB), (you, PRP), (l...</td>\n      <td>[\", how, do, you, like, them, apple, \", be, fr...</td>\n      <td>\" how do you like them apple \" be from the mov...</td>\n    </tr>\n  </tbody>\n</table>\n<p>10000 rows × 7 columns</p>\n</div>"
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Apply a lambda function to the 'lemmatized_text' column of the dataframe\n",
    "# This lambda function joins the list of lemmatized tokens into a single string\n",
    "# Each token is separated by a space\n",
    "df_tweet_sentiment[\"lemmatized_text_joined\"] = df_tweet_sentiment[\n",
    "    \"lemmatized_text\"\n",
    "].apply(lambda x: \" \".join(x))\n",
    "df_tweet_sentiment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:02.708266Z",
     "start_time": "2024-09-23T11:49:02.668342Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training texts: 8000\n",
      "Testing texts: 2000\n"
     ]
    }
   ],
   "source": [
    "x_train, x_test, y_train, y_test = train_test_split(\n",
    "    df_tweet_sentiment[\"lemmatized_text_joined\"],\n",
    "    df_tweet_sentiment[\"Sentiment\"],\n",
    "    test_size=0.2,\n",
    "    random_state=23828748,\n",
    ")\n",
    "print(\"Training texts:\", len(x_train))\n",
    "print(\"Testing texts:\", len(x_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vR52ysJYgjsI"
   },
   "source": [
    "# TF-Idf Vectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:04.273867Z",
     "start_time": "2024-09-23T11:49:04.243283Z"
    }
   },
   "outputs": [],
   "source": [
    "# Use unigram & bi-gram tf*idf features\n",
    "vectorizer = TfidfVectorizer(\n",
    "    ngram_range=(1, 2),\n",
    "    max_features=5000,\n",
    "    sublinear_tf=True,\n",
    "    stop_words=stopwords.words(\"english\"),\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:04.761596Z",
     "start_time": "2024-09-23T11:49:04.688198Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['abit' 'able' 'able get' 'able go' 'absolutely' 'abt' 'ac' 'accent'\n",
      " 'accident' 'accord']\n"
     ]
    }
   ],
   "source": [
    "# Fit Vectorizer on train data\n",
    "# Transform on all data (train - test)\n",
    "x_train_tfidf = vectorizer.fit_transform(x_train)\n",
    "x_test_tfidf = vectorizer.transform(x_test)\n",
    "\n",
    "print(vectorizer.get_feature_names_out()[:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:05.323170Z",
     "start_time": "2024-09-23T11:49:05.307419Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of training data: (8000, 5000)\n",
      "Shape of test data: (2000, 5000)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Shape of training data: {}\".format(x_train_tfidf.shape))\n",
    "print(\"Shape of test data: {}\\n\".format(x_test_tfidf.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:05.753319Z",
     "start_time": "2024-09-23T11:49:05.736471Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'scipy.sparse._csr.csr_matrix'>\n",
      "  (0, 981)\t0.2610531758549099\n",
      "  (0, 1751)\t0.25554243762248685\n",
      "  (0, 1828)\t0.5442627260277735\n",
      "  (0, 1973)\t0.19380063494342722\n",
      "  (0, 2777)\t0.3761954790174606\n",
      "  (0, 3286)\t0.3705922776538255\n",
      "  (0, 3364)\t0.33761772231426157\n",
      "  (0, 4306)\t0.29730987577023926\n",
      "  (0, 4637)\t0.17434063328693597\n",
      "  (0, 4648)\t0.14537359164051272\n"
     ]
    }
   ],
   "source": [
    "# sparse matrix --> Saves only the cells where tf*idf is not zero\n",
    "print(type(x_test_tfidf))\n",
    "print(x_test_tfidf[0, :])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "J-6kp8qzkHNl"
   },
   "source": [
    "# Baseline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:06.629666Z",
     "start_time": "2024-09-23T11:49:06.620525Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training took: 0.0008039474487304688 seconds \n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Train\n",
    "# The dummy classifier always predicts the 'most frequent' class\n",
    "baseline = DummyClassifier(strategy=\"most_frequent\")\n",
    "start_time = time.time()\n",
    "baseline.fit(x_train_tfidf, y_train)\n",
    "print(\"Training took: {} seconds \\n\".format(time.time() - start_time))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:06.987156Z",
     "start_time": "2024-09-23T11:49:06.961536Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train accuracy: 50.25%\n",
      "Test accuracy: 49.00% \n",
      "\n",
      "#Class 0 in Train dataset: (4020,)\n",
      "#Class 1 in Train dataset: (3980,)\n"
     ]
    }
   ],
   "source": [
    "# Evaluate\n",
    "predictions = baseline.predict(x_train_tfidf)\n",
    "score = accuracy_score(y_train, predictions)\n",
    "print(\"Train accuracy: {:.2f}%\".format(score * 100))\n",
    "\n",
    "predictions_test = baseline.predict(x_test_tfidf)\n",
    "score = accuracy_score(y_test, predictions_test)\n",
    "print(\"Test accuracy: {:.2f}% \\n\".format(score * 100))\n",
    "\n",
    "print(\"#Class 0 in Train dataset:\", np.where(y_train == 0)[0].shape)\n",
    "print(\"#Class 1 in Train dataset:\", np.where(y_train == 1)[0].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:07.318013Z",
     "start_time": "2024-09-23T11:49:07.291489Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report for TRAIN data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.50      1.00      0.67      4020\n",
      "           1       0.00      0.00      0.00      3980\n",
      "\n",
      "    accuracy                           0.50      8000\n",
      "   macro avg       0.25      0.50      0.33      8000\n",
      "weighted avg       0.25      0.50      0.34      8000\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/antonal/ta2/lib/python3.11/site-packages/sklearn/metrics/_classification.py:1517: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, f\"{metric.capitalize()} is\", len(result))\n",
      "/Users/antonal/ta2/lib/python3.11/site-packages/sklearn/metrics/_classification.py:1517: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, f\"{metric.capitalize()} is\", len(result))\n",
      "/Users/antonal/ta2/lib/python3.11/site-packages/sklearn/metrics/_classification.py:1517: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, f\"{metric.capitalize()} is\", len(result))\n"
     ]
    }
   ],
   "source": [
    "print(\"Classification Report for TRAIN data set:\")\n",
    "print(classification_report(y_train, predictions))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:07.917615Z",
     "start_time": "2024-09-23T11:49:07.893408Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report for TEST data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.49      1.00      0.66       980\n",
      "           1       0.00      0.00      0.00      1020\n",
      "\n",
      "    accuracy                           0.49      2000\n",
      "   macro avg       0.24      0.50      0.33      2000\n",
      "weighted avg       0.24      0.49      0.32      2000\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/antonal/ta2/lib/python3.11/site-packages/sklearn/metrics/_classification.py:1517: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, f\"{metric.capitalize()} is\", len(result))\n",
      "/Users/antonal/ta2/lib/python3.11/site-packages/sklearn/metrics/_classification.py:1517: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, f\"{metric.capitalize()} is\", len(result))\n",
      "/Users/antonal/ta2/lib/python3.11/site-packages/sklearn/metrics/_classification.py:1517: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, f\"{metric.capitalize()} is\", len(result))\n"
     ]
    }
   ],
   "source": [
    "print(\"Classification Report for TEST data set:\")\n",
    "print(classification_report(y_test, predictions_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:08.612118Z",
     "start_time": "2024-09-23T11:49:08.337002Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Confusion Matrix for test data:\n"
     ]
    },
    {
     "data": {
      "text/plain": "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x29dae81d0>"
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 2 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"\\nConfusion Matrix for test data:\")\n",
    "ConfusionMatrixDisplay.from_estimator(\n",
    "    baseline, x_test_tfidf, y_test, values_format=\".0f\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ipNyo53lkeKy"
   },
   "source": [
    "# Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:10.134954Z",
     "start_time": "2024-09-23T11:49:09.417720Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "0.752"
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Cross Validation Logistic Regression\n",
    "# Fit the model with CV\n",
    "log_cv = LogisticRegressionCV(\n",
    "    cv=3, Cs=8, random_state=2, penalty=\"l2\", solver=\"liblinear\"\n",
    ").fit(x_train_tfidf, y_train)\n",
    "# Score\n",
    "log_cv.score(x_test_tfidf, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:10.139415Z",
     "start_time": "2024-09-23T11:49:10.130999Z"
    }
   },
   "outputs": [],
   "source": [
    "predictions = log_cv.predict(x_train_tfidf)\n",
    "score = f1_score(y_train, predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:10.219076Z",
     "start_time": "2024-09-23T11:49:10.201022Z"
    }
   },
   "outputs": [],
   "source": [
    "predictions_test = log_cv.predict(x_test_tfidf)\n",
    "score_pred = f1_score(y_test, predictions_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:10.706815Z",
     "start_time": "2024-09-23T11:49:10.686406Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train f1-score: 89.71% \n",
      "\n",
      "Classification Report for TRAIN data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.90      0.90      0.90      4020\n",
      "           1       0.90      0.90      0.90      3980\n",
      "\n",
      "    accuracy                           0.90      8000\n",
      "   macro avg       0.90      0.90      0.90      8000\n",
      "weighted avg       0.90      0.90      0.90      8000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Train f1-score: {:.2f}% \\n\".format(score * 100))\n",
    "print(\"Classification Report for TRAIN data set:\")\n",
    "print(classification_report(y_train, predictions))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:11.177321Z",
     "start_time": "2024-09-23T11:49:11.141916Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test f1-score: 75.73% \n",
      "\n",
      "Classification Report for TEST data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.75      0.74      0.75       980\n",
      "           1       0.76      0.76      0.76      1020\n",
      "\n",
      "    accuracy                           0.75      2000\n",
      "   macro avg       0.75      0.75      0.75      2000\n",
      "weighted avg       0.75      0.75      0.75      2000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Test f1-score: {:.2f}% \\n\".format(score_pred * 100))\n",
    "print(\"Classification Report for TEST data set:\")\n",
    "print(classification_report(y_test, predictions_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 918
    },
    "id": "5jSvz2vMpAyF",
    "outputId": "2b1c353c-0731-47b8-b123-eb5cdccc2de0",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:11.667224Z",
     "start_time": "2024-09-23T11:49:11.594487Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Confusion Matrix for test data:\n"
     ]
    },
    {
     "data": {
      "text/plain": "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x29df6d310>"
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 2 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"\\nConfusion Matrix for test data:\")\n",
    "ConfusionMatrixDisplay.from_estimator(log_cv, x_test_tfidf, y_test, values_format=\".0f\")"
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "# MLP text classifier in Keras with centroids word embeddings"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "id": "KD_VNkezknMS",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:12.727884Z",
     "start_time": "2024-09-23T11:49:12.720220Z"
    }
   },
   "outputs": [],
   "source": [
    "X_train_new, X_val, y_train_new, y_val = train_test_split(\n",
    "    x_train, y_train, test_size=1 / 8, random_state=23828748\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "id": "ga8w4E4hFKuU",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:59:19.821061Z",
     "start_time": "2024-09-23T11:59:19.791289Z"
    }
   },
   "outputs": [],
   "source": [
    "# Convert with Keras\n",
    "y_train_1_hot = tf.keras.utils.to_categorical(\n",
    "    y_train_new, num_classes=len(y_train_new.unique())\n",
    ")  # 7000\n",
    "\n",
    "y_val_1_hot = tf.keras.utils.to_categorical(\n",
    "    y_val, num_classes=len(y_val.unique())\n",
    ")  # 1000\n",
    "\n",
    "y_test_1_hot = tf.keras.utils.to_categorical(\n",
    "    y_test, num_classes=len(y_test.unique())\n",
    ")  # 2000\n",
    "\n",
    "\n",
    "y_train_old_1_hot = tf.keras.utils.to_categorical(\n",
    "    y_train, num_classes=len(y_train.unique())\n",
    ")  # 8000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "2DuG0havoMS6",
    "outputId": "6a6d525a-40e9-4048-ac0c-a867b480b6f3",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:49.833151Z",
     "start_time": "2024-09-23T11:49:14.171924Z"
    }
   },
   "outputs": [],
   "source": [
    "# Download and load Google news embeddings\n",
    "wv = api.load(\"glove-twitter-100\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "BDJE0O3caG4c"
   },
   "source": [
    "## Calculate centroids"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "id": "LTr5Gs8Al5AT",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:49.837774Z",
     "start_time": "2024-09-23T11:49:49.834482Z"
    }
   },
   "outputs": [],
   "source": [
    "# Define a function to calculate the centroid of a text document\n",
    "def text_centroid(text, model):\n",
    "    text = text.split()  # Split the text into individual words\n",
    "    text_vec = []        # Initialize an empty list to store word vectors\n",
    "    counter = 0          # Counter to keep track of valid words (non-stopwords)\n",
    "\n",
    "    # Iterate through each word in the text\n",
    "    for word in text:\n",
    "        if word in STOP_WORDS:  # Skip stopwords (commonly used words that do not provide useful information)\n",
    "            continue\n",
    "        try:\n",
    "            if counter == 0:\n",
    "                # Get the word vector for the first valid word (in lowercase)\n",
    "                text_vec = model[word.lower()]\n",
    "            else:\n",
    "                # Add the word vector for the current word to the running sum\n",
    "                text_vec = np.add(text_vec, model[word.lower()])\n",
    "            counter += 1  # Increment the valid word counter\n",
    "        except:  # Skip words that are not in the model's vocabulary\n",
    "            pass\n",
    "\n",
    "    # If no valid words were found, return a zero vector\n",
    "    if counter == 0:\n",
    "        return np.zeros(model.vector_size)\n",
    "\n",
    "    # Return the average of the word vectors (centroid) by dividing the sum by the counter\n",
    "    return np.asarray(text_vec) / counter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "id": "Kds4AWAGqKtc",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:49:49.975322Z",
     "start_time": "2024-09-23T11:49:49.844491Z"
    }
   },
   "outputs": [],
   "source": [
    "# Calculate centroids for the training documents\n",
    "X_train_centroids = []\n",
    "for sentence in X_train_new:\n",
    "    X_train_centroids.append(text_centroid(sentence, wv))  # Calculate the centroid for each sentence\n",
    "X_train_centroids = np.stack(X_train_centroids, axis=0)    # Convert the list of centroids to a NumPy array\n",
    "\n",
    "# Calculate centroids for the validation documents\n",
    "X_val_centroids = []\n",
    "for sentence in X_val:\n",
    "    X_val_centroids.append(text_centroid(sentence, wv))     # Calculate centroids for validation data\n",
    "X_val_centroids = np.stack(X_val_centroids, axis=0)         # Convert to a NumPy array\n",
    "\n",
    "# Calculate centroids for the test documents\n",
    "X_test_centroids = []\n",
    "for sentence in x_test:\n",
    "    X_test_centroids.append(text_centroid(sentence, wv))    # Calculate centroids for test data\n",
    "X_test_centroids = np.stack(X_test_centroids, axis=0)       # Convert to a NumPy array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "outputs": [],
   "source": [
    "# Define a custom callback class Metrics that inherits from tf.keras.callbacks.Callback\n",
    "class Metrics(tf.keras.callbacks.Callback):\n",
    "    # Initialize the class with validation data\n",
    "    def __init__(self, valid_data):\n",
    "        # Call the parent class's initializer\n",
    "        super(Metrics, self).__init__()\n",
    "        # Store the validation data\n",
    "        self.validation_data = valid_data\n",
    "\n",
    "    # This method is called at the end of each epoch\n",
    "    def on_epoch_end(self, epoch, logs=None):\n",
    "        # If logs dictionary is not provided, initialize it as an empty dictionary\n",
    "        logs = logs or {}\n",
    "        # Predict the output for validation data and take the class with the highest probability\n",
    "        val_predict = np.argmax(self.model.predict(self.validation_data[0]), -1)\n",
    "        # Get the actual validation targets\n",
    "        val_targ = self.validation_data[1]\n",
    "        # Cast the validation targets to float32 type\n",
    "        val_targ = tf.cast(val_targ, dtype=tf.float32)\n",
    "        # If the validation targets are one-hot encoded, convert them to label encoded\n",
    "        if len(val_targ.shape) == 2 and val_targ.shape[1] != 1:\n",
    "            val_targ = np.argmax(val_targ, -1)\n",
    "\n",
    "        # Calculate the weighted F1 score\n",
    "        _val_f1 = f1_score(val_targ, val_predict, average=\"weighted\")\n",
    "        # Calculate the weighted recall\n",
    "        _val_recall = recall_score(val_targ, val_predict, average=\"weighted\")\n",
    "        # Calculate the weighted precision\n",
    "        _val_precision = precision_score(val_targ, val_predict, average=\"weighted\")\n",
    "\n",
    "        # Log the calculated metrics\n",
    "        logs[\"val_f1\"] = _val_f1\n",
    "        logs[\"val_recall\"] = _val_recall\n",
    "        logs[\"val_precision\"] = _val_precision\n",
    "        # Print the calculated metrics\n",
    "        print(\n",
    "            \" — val_f1: %f — val_precision: %f — val_recall: %f\"\n",
    "            % (_val_f1, _val_precision, _val_recall)\n",
    "        )\n",
    "        return"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:51:02.822666Z",
     "start_time": "2024-09-23T11:51:02.800576Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "voqKW_NnZhoC",
    "outputId": "4788ec83-c9ab-411b-80dd-ec9e1a58fa34",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:51:31.205084Z",
     "start_time": "2024-09-23T11:51:16.842557Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/antonal/ta2/lib/python3.11/site-packages/keras/src/layers/core/dense.py:87: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.\n",
      "  super().__init__(activity_regularizer=activity_regularizer, **kwargs)\n"
     ]
    },
    {
     "data": {
      "text/plain": "\u001B[1mModel: \"sequential_1\"\u001B[0m\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"sequential_1\"</span>\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n┃\u001B[1m \u001B[0m\u001B[1mLayer (type)                   \u001B[0m\u001B[1m \u001B[0m┃\u001B[1m \u001B[0m\u001B[1mOutput Shape          \u001B[0m\u001B[1m \u001B[0m┃\u001B[1m \u001B[0m\u001B[1m      Param #\u001B[0m\u001B[1m \u001B[0m┃\n┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n│ dense_4 (\u001B[38;5;33mDense\u001B[0m)                 │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m1024\u001B[0m)           │       \u001B[38;5;34m103,424\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_3 (\u001B[38;5;33mDropout\u001B[0m)             │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m1024\u001B[0m)           │             \u001B[38;5;34m0\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_5 (\u001B[38;5;33mDense\u001B[0m)                 │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m1024\u001B[0m)           │     \u001B[38;5;34m1,049,600\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_4 (\u001B[38;5;33mDropout\u001B[0m)             │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m1024\u001B[0m)           │             \u001B[38;5;34m0\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_6 (\u001B[38;5;33mDense\u001B[0m)                 │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m512\u001B[0m)            │       \u001B[38;5;34m524,800\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_5 (\u001B[38;5;33mDropout\u001B[0m)             │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m512\u001B[0m)            │             \u001B[38;5;34m0\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_7 (\u001B[38;5;33mDense\u001B[0m)                 │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m2\u001B[0m)              │         \u001B[38;5;34m1,026\u001B[0m │\n└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n┃<span style=\"font-weight: bold\"> Layer (type)                    </span>┃<span style=\"font-weight: bold\"> Output Shape           </span>┃<span style=\"font-weight: bold\">       Param # </span>┃\n┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n│ dense_4 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1024</span>)           │       <span style=\"color: #00af00; text-decoration-color: #00af00\">103,424</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_3 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)             │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1024</span>)           │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_5 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1024</span>)           │     <span style=\"color: #00af00; text-decoration-color: #00af00\">1,049,600</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_4 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)             │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1024</span>)           │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_6 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">512</span>)            │       <span style=\"color: #00af00; text-decoration-color: #00af00\">524,800</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_5 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)             │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">512</span>)            │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_7 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">2</span>)              │         <span style=\"color: #00af00; text-decoration-color: #00af00\">1,026</span> │\n└─────────────────────────────────┴────────────────────────┴───────────────┘\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "\u001B[1m Total params: \u001B[0m\u001B[38;5;34m1,678,850\u001B[0m (6.40 MB)\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">1,678,850</span> (6.40 MB)\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "\u001B[1m Trainable params: \u001B[0m\u001B[38;5;34m1,678,850\u001B[0m (6.40 MB)\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">1,678,850</span> (6.40 MB)\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "\u001B[1m Non-trainable params: \u001B[0m\u001B[38;5;34m0\u001B[0m (0.00 B)\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "None\n",
      "Epoch 1/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 2ms/step\n",
      " — val_f1: 0.731704 — val_precision: 0.749755 — val_recall: 0.734000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m2s\u001B[0m 37ms/step - categorical_accuracy: 0.6159 - loss: 0.6419 - val_categorical_accuracy: 0.7340 - val_loss: 0.5251 - val_f1: 0.7317 - val_recall: 0.7340 - val_precision: 0.7498\n",
      "Epoch 2/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.746784 — val_precision: 0.751444 — val_recall: 0.747000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 21ms/step - categorical_accuracy: 0.7637 - loss: 0.4929 - val_categorical_accuracy: 0.7470 - val_loss: 0.4896 - val_f1: 0.7468 - val_recall: 0.7470 - val_precision: 0.7514\n",
      "Epoch 3/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.741275 — val_precision: 0.749680 — val_recall: 0.742000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.7749 - loss: 0.4640 - val_categorical_accuracy: 0.7420 - val_loss: 0.4960 - val_f1: 0.7413 - val_recall: 0.7420 - val_precision: 0.7497\n",
      "Epoch 4/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.758943 — val_precision: 0.758924 — val_recall: 0.759000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.7891 - loss: 0.4541 - val_categorical_accuracy: 0.7590 - val_loss: 0.4758 - val_f1: 0.7589 - val_recall: 0.7590 - val_precision: 0.7589\n",
      "Epoch 5/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.766101 — val_precision: 0.767085 — val_recall: 0.766000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.7968 - loss: 0.4296 - val_categorical_accuracy: 0.7660 - val_loss: 0.4905 - val_f1: 0.7661 - val_recall: 0.7660 - val_precision: 0.7671\n",
      "Epoch 6/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.759401 — val_precision: 0.763962 — val_recall: 0.761000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 19ms/step - categorical_accuracy: 0.8092 - loss: 0.4167 - val_categorical_accuracy: 0.7610 - val_loss: 0.4762 - val_f1: 0.7594 - val_recall: 0.7610 - val_precision: 0.7640\n",
      "Epoch 7/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.771100 — val_precision: 0.772016 — val_recall: 0.771000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 17ms/step - categorical_accuracy: 0.8043 - loss: 0.4104 - val_categorical_accuracy: 0.7710 - val_loss: 0.4902 - val_f1: 0.7711 - val_recall: 0.7710 - val_precision: 0.7720\n",
      "Epoch 8/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.769402 — val_precision: 0.770599 — val_recall: 0.770000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 17ms/step - categorical_accuracy: 0.8268 - loss: 0.3829 - val_categorical_accuracy: 0.7700 - val_loss: 0.4841 - val_f1: 0.7694 - val_recall: 0.7700 - val_precision: 0.7706\n",
      "Epoch 9/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 992us/step- categ\n",
      " — val_f1: 0.766714 — val_precision: 0.770326 — val_recall: 0.768000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.8377 - loss: 0.3652 - val_categorical_accuracy: 0.7680 - val_loss: 0.5040 - val_f1: 0.7667 - val_recall: 0.7680 - val_precision: 0.7703\n",
      "Epoch 10/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.761071 — val_precision: 0.761265 — val_recall: 0.761000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 19ms/step - categorical_accuracy: 0.8430 - loss: 0.3563 - val_categorical_accuracy: 0.7610 - val_loss: 0.5150 - val_f1: 0.7611 - val_recall: 0.7610 - val_precision: 0.7613\n",
      "Epoch 11/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.761082 — val_precision: 0.761347 — val_recall: 0.761000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.8644 - loss: 0.3273 - val_categorical_accuracy: 0.7610 - val_loss: 0.5560 - val_f1: 0.7611 - val_recall: 0.7610 - val_precision: 0.7613\n",
      "Epoch 12/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.752079 — val_precision: 0.753700 — val_recall: 0.752000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.8659 - loss: 0.3035 - val_categorical_accuracy: 0.7520 - val_loss: 0.5643 - val_f1: 0.7521 - val_recall: 0.7520 - val_precision: 0.7537\n",
      "Epoch 13/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.758967 — val_precision: 0.758948 — val_recall: 0.759000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 17ms/step - categorical_accuracy: 0.8712 - loss: 0.3058 - val_categorical_accuracy: 0.7590 - val_loss: 0.6054 - val_f1: 0.7590 - val_recall: 0.7590 - val_precision: 0.7589\n",
      "Epoch 14/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.754916 — val_precision: 0.754906 — val_recall: 0.755000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 18ms/step - categorical_accuracy: 0.8831 - loss: 0.2688 - val_categorical_accuracy: 0.7550 - val_loss: 0.5798 - val_f1: 0.7549 - val_recall: 0.7550 - val_precision: 0.7549\n",
      "Epoch 15/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.765906 — val_precision: 0.769655 — val_recall: 0.766000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.8903 - loss: 0.2571 - val_categorical_accuracy: 0.7660 - val_loss: 0.5995 - val_f1: 0.7659 - val_recall: 0.7660 - val_precision: 0.7697\n",
      "Epoch 16/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.765044 — val_precision: 0.765125 — val_recall: 0.765000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.8906 - loss: 0.2585 - val_categorical_accuracy: 0.7650 - val_loss: 0.6426 - val_f1: 0.7650 - val_recall: 0.7650 - val_precision: 0.7651\n",
      "Epoch 17/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.751692 — val_precision: 0.757201 — val_recall: 0.752000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.9071 - loss: 0.2257 - val_categorical_accuracy: 0.7520 - val_loss: 0.7155 - val_f1: 0.7517 - val_recall: 0.7520 - val_precision: 0.7572\n",
      "Epoch 18/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.767496 — val_precision: 0.768384 — val_recall: 0.768000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 19ms/step - categorical_accuracy: 0.9126 - loss: 0.2167 - val_categorical_accuracy: 0.7680 - val_loss: 0.7192 - val_f1: 0.7675 - val_recall: 0.7680 - val_precision: 0.7684\n",
      "Epoch 19/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.742021 — val_precision: 0.742047 — val_recall: 0.742000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.9127 - loss: 0.2049 - val_categorical_accuracy: 0.7420 - val_loss: 0.8005 - val_f1: 0.7420 - val_recall: 0.7420 - val_precision: 0.7420\n",
      "Epoch 20/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.753898 — val_precision: 0.756469 — val_recall: 0.755000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 20ms/step - categorical_accuracy: 0.9208 - loss: 0.1916 - val_categorical_accuracy: 0.7550 - val_loss: 0.7439 - val_f1: 0.7539 - val_recall: 0.7550 - val_precision: 0.7565\n",
      "Epoch 21/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.742093 — val_precision: 0.742400 — val_recall: 0.742000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 18ms/step - categorical_accuracy: 0.9263 - loss: 0.1776 - val_categorical_accuracy: 0.7420 - val_loss: 0.8531 - val_f1: 0.7421 - val_recall: 0.7420 - val_precision: 0.7424\n",
      "Epoch 22/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.760000 — val_precision: 0.760000 — val_recall: 0.760000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 19ms/step - categorical_accuracy: 0.9328 - loss: 0.1719 - val_categorical_accuracy: 0.7600 - val_loss: 0.9715 - val_f1: 0.7600 - val_recall: 0.7600 - val_precision: 0.7600\n",
      "Epoch 23/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 2ms/step\n",
      " — val_f1: 0.752097 — val_precision: 0.753374 — val_recall: 0.752000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 21ms/step - categorical_accuracy: 0.9332 - loss: 0.1727 - val_categorical_accuracy: 0.7520 - val_loss: 0.8778 - val_f1: 0.7521 - val_recall: 0.7520 - val_precision: 0.7534\n",
      "Epoch 24/100\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/stepp - categ\n",
      " — val_f1: 0.743011 — val_precision: 0.743023 — val_recall: 0.743000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m1s\u001B[0m 27ms/step - categorical_accuracy: 0.9393 - loss: 0.1518 - val_categorical_accuracy: 0.7430 - val_loss: 0.8102 - val_f1: 0.7430 - val_recall: 0.7430 - val_precision: 0.7430\n",
      "Epoch 24: early stopping\n",
      "Restoring model weights from the end of the best epoch: 4.\n"
     ]
    }
   ],
   "source": [
    "# Define a new Sequential model\n",
    "model2 = Sequential()\n",
    "\n",
    "# Add the first dense (fully connected) layer with 1024 units and ReLU activation\n",
    "# The input dimension is set to the number of features in the centroid vectors (X_train_centroids)\n",
    "model2.add(Dense(1024, input_dim=X_train_centroids.shape[1], activation=\"relu\"))\n",
    "\n",
    "# Add a Dropout layer to reduce overfitting with a dropout rate of 0.3 (30% of neurons dropped)\n",
    "model2.add(Dropout(0.3))\n",
    "\n",
    "# Add a second dense layer with 1024 units and ReLU activation\n",
    "model2.add(Dense(1024, activation=\"relu\"))\n",
    "\n",
    "# Add another Dropout layer with a 30% dropout rate\n",
    "model2.add(Dropout(0.3))\n",
    "\n",
    "# Add a third dense layer with 512 units and ReLU activation\n",
    "model2.add(Dense(512, activation=\"relu\"))\n",
    "\n",
    "# Add another Dropout layer with a 30% dropout rate\n",
    "model2.add(Dropout(0.3))\n",
    "\n",
    "# Add the output layer with a softmax activation to classify into the number of unique classes\n",
    "# The number of units in the output layer matches the number of unique classes in the target\n",
    "model2.add(Dense(len(y_train_new.unique()), activation=\"softmax\"))\n",
    "\n",
    "# Print the model summary to display the structure and layers of the model\n",
    "print(model2.summary())\n",
    "\n",
    "# Compile the model with categorical cross-entropy as the loss function, Adam optimizer, and categorical accuracy metric\n",
    "model2.compile(\n",
    "    loss=\"categorical_crossentropy\",            # Loss function for multi-class classification\n",
    "    optimizer=Adam(learning_rate=0.001),        # Adam optimizer with a learning rate of 0.001\n",
    "    metrics=[CategoricalAccuracy()],            # Use categorical accuracy as the evaluation metric\n",
    ")\n",
    "\n",
    "# Define an EarlyStopping callback to stop training if the validation loss doesn't improve for 20 epochs\n",
    "early_stopping = EarlyStopping(patience=20, verbose=2, restore_best_weights=True)\n",
    "\n",
    "# Train the model using the training centroid vectors and one-hot encoded labels\n",
    "history2 = model2.fit(\n",
    "    X_train_centroids,               # Training feature set (centroid vectors)\n",
    "    y_train_1_hot,                   # Training labels (one-hot encoded)\n",
    "    validation_data=(X_val_centroids, y_val_1_hot),  # Validation data (centroid vectors and labels)\n",
    "    batch_size=256,                  # Batch size of 256\n",
    "    epochs=100,                      # Maximum number of epochs to train\n",
    "    shuffle=True,                    # Shuffle the training data at the start of each epoch\n",
    "    callbacks=[\n",
    "        Metrics(valid_data=(X_val_centroids, y_val_1_hot)),  # Custom metric callback for monitoring performance\n",
    "        early_stopping,  # Apply early stopping to prevent overfitting and restore the best weights\n",
    "    ],\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "h74dNgiVbQJE"
   },
   "source": [
    "# Visualize Model's Training History"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:52:03.938899Z",
     "start_time": "2024-09-23T11:52:03.856815Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for accuracy\n",
    "plt.plot(history2.history[\"categorical_accuracy\"])\n",
    "plt.plot(history2.history[\"val_categorical_accuracy\"])\n",
    "plt.title(\"model accuracy\")\n",
    "plt.ylabel(\"accuracy\")\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.legend([\"train\", \"dev\"], loc=\"upper left\")\n",
    "plt.xticks(range(1, len(history2.history[\"categorical_accuracy\"]) + 1, 4))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 927
    },
    "id": "RcYQpGqfZhq-",
    "outputId": "82bd2761-0df4-48bc-b13f-8e4ab80139ae",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:52:04.855865Z",
     "start_time": "2024-09-23T11:52:04.770058Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for loss\n",
    "plt.plot(history2.history[\"loss\"])\n",
    "plt.plot(history2.history[\"val_loss\"])\n",
    "plt.title(\"model loss\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.legend([\"train\", \"dev\"], loc=\"upper right\")\n",
    "plt.xticks(range(1, len(history2.history[\"loss\"]) + 1, 4))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "id": "a8fuanqDZhtu",
    "ExecuteTime": {
     "end_time": "2024-09-23T11:52:40.614065Z",
     "start_time": "2024-09-23T11:52:40.501118Z"
    }
   },
   "outputs": [],
   "source": [
    "# create checkpoint dir if not exists\n",
    "if not os.path.exists(\"checkpoints\"):\n",
    "    os.makedirs(\"checkpoints\")\n",
    "\n",
    "# Save model & weights\n",
    "model2.save(\"checkpoints/MLP_centroids.keras\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "outputs": [],
   "source": [
    "# load weights and model from the saved checkpoint\n",
    "centroids_model = load_model(\"checkpoints/MLP_centroids.keras\")"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:52:50.785961Z",
     "start_time": "2024-09-23T11:52:50.714199Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001B[1m63/63\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m0s\u001B[0m 1ms/step\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.77      0.79      0.78       980\n",
      "           1       0.79      0.77      0.78      1020\n",
      "\n",
      "    accuracy                           0.78      2000\n",
      "   macro avg       0.78      0.78      0.78      2000\n",
      "weighted avg       0.78      0.78      0.78      2000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Predict the class labels for the test data using model2 (centroid-based model)\n",
    "predictions = np.argmax(model2.predict(X_test_centroids), -1)\n",
    "\n",
    "# Print a classification report comparing the true labels and predicted labels for the test set\n",
    "print(classification_report(y_test, predictions))"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T11:52:52.716510Z",
     "start_time": "2024-09-23T11:52:52.568914Z"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# RNN GRU text classifier\n",
    "\n",
    "We need to tokenize our dataset so we can use the tokenized data in order to build the padding and the embedded vocabulary that we will use in the RNN GRU"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "outputs": [],
   "source": [
    "def is_number(s):\n",
    "    # Try to convert the input to a float\n",
    "    try:\n",
    "        float(s)\n",
    "        return True\n",
    "    except ValueError:\n",
    "        # If a ValueError is raised, the input is not a number\n",
    "        return False"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:08:01.762385Z",
     "start_time": "2024-09-23T12:08:01.750862Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "outputs": [
    {
     "data": {
      "text/plain": "<spacy.pipeline.sentencizer.Sentencizer at 0x2ffeddb50>"
     },
     "execution_count": 87,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load the English language model from Spacy, disabling the tagger, parser, named entity recognizer, and lemmatizer components\n",
    "nlp = spacy.load(\"en_core_web_sm\", disable=[\"tagger\", \"parser\",\"ner\", \"lemmatizer\"])\n",
    "\n",
    "# Add the sentencizer component to the pipeline. This component is responsible for splitting the text into sentences\n",
    "nlp.add_pipe('sentencizer')"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:08:02.820160Z",
     "start_time": "2024-09-23T12:08:02.543763Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 7000/7000 [00:12<00:00, 563.90it/s]\n"
     ]
    }
   ],
   "source": [
    "# Initialize an empty list to store the tokenized training data\n",
    "X_train_tokenized = []\n",
    "\n",
    "# Loop over each item in the training data\n",
    "for idx in tqdm(range(len(X_train_new))):\n",
    "  # Parse the text with Spacy\n",
    "  doc = nlp(X_train_new.iloc[idx])\n",
    "  # Initialize an empty list to store the tokens for this document\n",
    "  tokens = []\n",
    "  # Loop over each sentence in the document\n",
    "  for sent in doc.sents:\n",
    "    # Loop over each token in the sentence\n",
    "    for tok in sent:\n",
    "      # If the token is a newline, tab, double dash, asterisk, stop word, punctuation, all punctuation, or a number, skip it\n",
    "      if '\\n' in tok.text or \"\\t\" in tok.text or \"--\" in tok.text or \"*\" in tok.text or\\\n",
    "       tok.text.lower() in STOP_WORDS or tok.text in string.punctuation or\\\n",
    "        all(x in string.punctuation for x in tok.text) or is_number(tok.text):\n",
    "        continue\n",
    "      # If the token is not empty, add it to the list of tokens for this document\n",
    "      if tok.text.strip():\n",
    "        tokens.append(tok.text.replace('\"',\"'\").strip().lower())\n",
    "  # Add the list of tokens for this document to the list of tokenized training data\n",
    "  X_train_tokenized.append(tokens)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:10:55.396940Z",
     "start_time": "2024-09-23T12:10:42.973852Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1000/1000 [00:01<00:00, 524.42it/s]\n"
     ]
    }
   ],
   "source": [
    "# Loop over each item in the training data\n",
    "X_val_tokenized = []\n",
    "\n",
    "# Loop over each item in the training data\n",
    "for idx in tqdm(range(len(X_val))):\n",
    "  # Parse the text with Spacy\n",
    "  doc = nlp(X_val.iloc[idx])\n",
    "  # Initialize an empty list to store the tokens for this document\n",
    "  tokens = []\n",
    "  # Loop over each sentence in the document\n",
    "  for sent in doc.sents:\n",
    "    # Loop over each token in the sentence\n",
    "    for tok in sent:\n",
    "      # If the token is a newline, tab, double dash, asterisk, stop word, punctuation, all punctuation, or a number, skip it\n",
    "      if '\\n' in tok.text or \"\\t\" in tok.text or \"--\" in tok.text or \"*\" in tok.text or\\\n",
    "       tok.text.lower() in STOP_WORDS or tok.text in string.punctuation or\\\n",
    "        all(x in string.punctuation for x in tok.text) or is_number(tok.text):\n",
    "        continue\n",
    "      # If the token is not empty, add it to the list of tokens for this document\n",
    "      if tok.text.strip():\n",
    "        tokens.append(tok.text.replace('\"',\"'\").strip().lower())\n",
    "  # Add the list of tokens for this document to the list of tokenized validation data\n",
    "  X_val_tokenized.append(tokens)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:08:17.822309Z",
     "start_time": "2024-09-23T12:08:15.909994Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2000/2000 [00:03<00:00, 517.96it/s]\n"
     ]
    }
   ],
   "source": [
    "X_test_tokenized = []\n",
    "\n",
    "# Loop over each item in the training data\n",
    "for idx in tqdm(range(len(x_test))):\n",
    "  # Parse the text with Spacy\n",
    "  doc = nlp(x_test.iloc[idx])\n",
    "  # Initialize an empty list to store the tokens for this document\n",
    "  tokens = []\n",
    "  # Loop over each sentence in the document\n",
    "  for sent in doc.sents:\n",
    "    # Loop over each token in the sentence\n",
    "    for tok in sent:\n",
    "      # If the token is a newline, tab, double dash, asterisk, stop word, punctuation, all punctuation, or a number, skip it\n",
    "      if '\\n' in tok.text or \"\\t\" in tok.text or \"--\" in tok.text or \"*\" in tok.text or\\\n",
    "       tok.text.lower() in STOP_WORDS or tok.text in string.punctuation or\\\n",
    "        all(x in string.punctuation for x in tok.text) or is_number(tok.text):\n",
    "        continue\n",
    "      # If the token is not empty, add it to the list of tokens for this document\n",
    "      if tok.text.strip():\n",
    "        tokens.append(tok.text.replace('\"',\"'\").strip().lower())\n",
    "  # Add the list of tokens for this document to the list of tokenized testing data\n",
    "  X_test_tokenized.append(tokens)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:08:21.695319Z",
     "start_time": "2024-09-23T12:08:17.821853Z"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Vocabulary Creation"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "outputs": [],
   "source": [
    "# Define the maximum number of words to keep, based on word frequency\n",
    "MAX_WORDS = 512\n",
    "\n",
    "# Define the maximum length of each sequence\n",
    "MAX_SEQUENCE_LENGTH = 256\n",
    "\n",
    "# Define the dimension of the dense embedding\n",
    "EMBEDDING_DIM = 100"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:08:21.695400Z",
     "start_time": "2024-09-23T12:08:21.687064Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(20, 'feel')\n",
      "(1, 'shallow')\n",
      "(1, 'breath')\n",
      "(1, 'squeeze')\n",
      "(1, 'past')\n",
      "(1, 'pound')\n",
      "(1, 'stone')\n",
      "(1, 'laid')\n",
      "(1, 'chest')\n"
     ]
    }
   ],
   "source": [
    "# Create a TextVectorization layer\n",
    "# This layer transforms strings into integer representations that can be read by an Embedding layer\n",
    "vectorizer = TextVectorization(max_tokens=MAX_WORDS, output_mode='int', ngrams=1,\n",
    "              output_sequence_length=MAX_SEQUENCE_LENGTH)\n",
    "\n",
    "# Adapt the vectorizer to the text data\n",
    "vectorizer.adapt([\" \".join(x) for x in X_train_tokenized])\n",
    "\n",
    "# Use the vectorizer to transform the 15th document in X_train_tokenized into integer representations\n",
    "# Then print each word in the document along with its corresponding integer representation\n",
    "for widx, word in zip(vectorizer(' '.join(X_train_tokenized[14])).numpy().tolist(),X_train_tokenized[14]):\n",
    "    print((widx,word))"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:14:32.927595Z",
     "start_time": "2024-09-23T12:14:32.870291Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "outputs": [],
   "source": [
    "# Initialize an embedding matrix with zeros\n",
    "# The number of rows is equal to the maximum number of words in the vocabulary (MAX_WORDS)\n",
    "# The number of columns is equal to the dimension of the word vectors (100 in this case)\n",
    "embedding_matrix = np.zeros(shape=(MAX_WORDS, 100))\n",
    "\n",
    "# Enumerate over the vocabulary obtained from the vectorizer\n",
    "# w2idx is the index of the word in the vocabulary\n",
    "# _word is the actual word string\n",
    "for w2idx, _word in enumerate(vectorizer.get_vocabulary()):\n",
    "    # Skip the first two tokens ('PAD' and 'UNK') as they do not have corresponding word vectors\n",
    "    if w2idx < 2:\n",
    "      continue\n",
    "    try:\n",
    "      # Try to get the word vector for the current word from the word2vec model (wv)\n",
    "      # Assign this word vector to the corresponding row in the embedding matrix\n",
    "      embedding_matrix[w2idx] = wv[_word]\n",
    "    except KeyError:\n",
    "      # If the word is not found in the word2vec model, catch the KeyError and continue to the next word\n",
    "      continue"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:24:33.304718Z",
     "start_time": "2024-09-23T12:24:33.265759Z"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "### BiGRU and deep self-attention"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "outputs": [],
   "source": [
    "class SelfAttention(Layer):\n",
    "  def __init__(self, mlp_layers=0, units=0, dropout_rate=0, return_attention=False, **kwargs):\n",
    "    # Initialize the parent class\n",
    "    super(SelfAttention, self).__init__(**kwargs)\n",
    "\n",
    "    # Store the parameters\n",
    "    self.mlp_layers = mlp_layers\n",
    "    self.mlp_units = units\n",
    "    self.return_attention = return_attention\n",
    "    self.dropout_rate = dropout_rate\n",
    "\n",
    "    # Build the MLP (Multi-Layer Perceptron) that will be used to compute the attention scores\n",
    "    self.attention_mlp = self.build_mlp()\n",
    "\n",
    "  def build_mlp(self):\n",
    "    # Initialize the MLP\n",
    "    mlp = Sequential()\n",
    "\n",
    "    # Add the hidden layers\n",
    "    for i in range(self.mlp_layers):\n",
    "      mlp.add(Dense(self.mlp_units, activation='relu'))\n",
    "      mlp.add(Dropout(self.dropout_rate))\n",
    "\n",
    "    # Add the output layer\n",
    "    mlp.add(Dense(1))\n",
    "\n",
    "    return mlp\n",
    "\n",
    "  def call(self, x, mask=None):\n",
    "    # Compute the attention scores\n",
    "    a = self.attention_mlp(x)\n",
    "    a = tf.squeeze(a, axis=2)\n",
    "\n",
    "    # Apply the mask if provided\n",
    "    if mask is not None:\n",
    "      mask = tf.keras.backend.cast(mask, tf.keras.backend.floatx())\n",
    "      a -= 100000.0 * (1.0 - mask)\n",
    "\n",
    "    # Compute the softmax of the attention scores\n",
    "    a = tf.keras.backend.expand_dims(tf.keras.backend.softmax(a, axis=-1))\n",
    "\n",
    "    # Compute the weighted sum of the inputs according to the attention scores\n",
    "    weighted_input = x * a\n",
    "    result = tf.keras.backend.sum(weighted_input, axis=1)\n",
    "\n",
    "    # Return the result and the attention scores if required\n",
    "    if self.return_attention:\n",
    "      return [result, a]\n",
    "    return result"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:24:34.595689Z",
     "start_time": "2024-09-23T12:24:34.589574Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "outputs": [],
   "source": [
    "# Define the size of the GRU layer\n",
    "GRU_SIZE = 300\n",
    "\n",
    "# Define the size of the dense layer\n",
    "DENSE = 1000"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T12:24:35.685006Z",
     "start_time": "2024-09-23T12:24:35.656763Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/antonal/ta2/lib/python3.11/site-packages/keras/src/layers/core/embedding.py:90: UserWarning: Argument `input_length` is deprecated. Just remove it.\n",
      "  warnings.warn(\n",
      "/Users/antonal/ta2/lib/python3.11/site-packages/keras/src/layers/layer.py:934: UserWarning: Layer 'self_attention_10' (of type SelfAttention) was passed an input with a mask attached to it. However, this layer does not support masking and will therefore destroy the mask information. Downstream layers will not see the mask.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "text/plain": "\u001B[1mModel: \"sequential_23\"\u001B[0m\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"sequential_23\"</span>\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n┃\u001B[1m \u001B[0m\u001B[1mLayer (type)                   \u001B[0m\u001B[1m \u001B[0m┃\u001B[1m \u001B[0m\u001B[1mOutput Shape          \u001B[0m\u001B[1m \u001B[0m┃\u001B[1m \u001B[0m\u001B[1m      Param #\u001B[0m\u001B[1m \u001B[0m┃\n┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n│ text_vectorization_4            │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m256\u001B[0m)            │             \u001B[38;5;34m0\u001B[0m │\n│ (\u001B[38;5;33mTextVectorization\u001B[0m)             │                        │               │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ embedding_11 (\u001B[38;5;33mEmbedding\u001B[0m)        │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m256\u001B[0m, \u001B[38;5;34m100\u001B[0m)       │        \u001B[38;5;34m51,200\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_47 (\u001B[38;5;33mDropout\u001B[0m)            │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m256\u001B[0m, \u001B[38;5;34m100\u001B[0m)       │             \u001B[38;5;34m0\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ bidirectional_10                │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m256\u001B[0m, \u001B[38;5;34m600\u001B[0m)       │       \u001B[38;5;34m723,600\u001B[0m │\n│ (\u001B[38;5;33mBidirectional\u001B[0m)                 │                        │               │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ self_attention_10               │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m600\u001B[0m)            │       \u001B[38;5;34m721,801\u001B[0m │\n│ (\u001B[38;5;33mSelfAttention\u001B[0m)                 │                        │               │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_61 (\u001B[38;5;33mDense\u001B[0m)                │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m1000\u001B[0m)           │       \u001B[38;5;34m601,000\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_50 (\u001B[38;5;33mDropout\u001B[0m)            │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m1000\u001B[0m)           │             \u001B[38;5;34m0\u001B[0m │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_62 (\u001B[38;5;33mDense\u001B[0m)                │ (\u001B[38;5;45mNone\u001B[0m, \u001B[38;5;34m2\u001B[0m)              │         \u001B[38;5;34m2,002\u001B[0m │\n└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n┃<span style=\"font-weight: bold\"> Layer (type)                    </span>┃<span style=\"font-weight: bold\"> Output Shape           </span>┃<span style=\"font-weight: bold\">       Param # </span>┃\n┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n│ text_vectorization_4            │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>)            │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">TextVectorization</span>)             │                        │               │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ embedding_11 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Embedding</span>)        │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>)       │        <span style=\"color: #00af00; text-decoration-color: #00af00\">51,200</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_47 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)            │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>)       │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ bidirectional_10                │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">600</span>)       │       <span style=\"color: #00af00; text-decoration-color: #00af00\">723,600</span> │\n│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Bidirectional</span>)                 │                        │               │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ self_attention_10               │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">600</span>)            │       <span style=\"color: #00af00; text-decoration-color: #00af00\">721,801</span> │\n│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">SelfAttention</span>)                 │                        │               │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_61 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1000</span>)           │       <span style=\"color: #00af00; text-decoration-color: #00af00\">601,000</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dropout_50 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)            │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1000</span>)           │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n├─────────────────────────────────┼────────────────────────┼───────────────┤\n│ dense_62 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">2</span>)              │         <span style=\"color: #00af00; text-decoration-color: #00af00\">2,002</span> │\n└─────────────────────────────────┴────────────────────────┴───────────────┘\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "\u001B[1m Total params: \u001B[0m\u001B[38;5;34m2,099,603\u001B[0m (8.01 MB)\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">2,099,603</span> (8.01 MB)\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "\u001B[1m Trainable params: \u001B[0m\u001B[38;5;34m2,048,403\u001B[0m (7.81 MB)\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">2,048,403</span> (7.81 MB)\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "\u001B[1m Non-trainable params: \u001B[0m\u001B[38;5;34m51,200\u001B[0m (200.00 KB)\n",
      "text/html": "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">51,200</span> (200.00 KB)\n</pre>\n"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "None\n",
      "Epoch 1/20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/antonal/ta2/lib/python3.11/site-packages/keras/src/layers/layer.py:934: UserWarning: Layer 'sequential_24' (of type Sequential) was passed an input with a mask attached to it. However, this layer does not support masking and will therefore destroy the mask information. Downstream layers will not see the mask.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m9s\u001B[0m 272ms/step\n",
      " — val_f1: 0.711949 — val_precision: 0.721342 — val_recall: 0.713000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m100s\u001B[0m 3s/step - categorical_accuracy: 0.6408 - loss: 0.6298 - val_categorical_accuracy: 0.7130 - val_loss: 0.5461 - val_f1: 0.7119 - val_recall: 0.7130 - val_precision: 0.7213\n",
      "Epoch 2/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m20s\u001B[0m 627ms/step\n",
      " — val_f1: 0.706201 — val_precision: 0.724831 — val_recall: 0.709000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m144s\u001B[0m 5s/step - categorical_accuracy: 0.7245 - loss: 0.5461 - val_categorical_accuracy: 0.7090 - val_loss: 0.5606 - val_f1: 0.7062 - val_recall: 0.7090 - val_precision: 0.7248\n",
      "Epoch 3/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m10s\u001B[0m 318ms/step\n",
      " — val_f1: 0.736651 — val_precision: 0.736954 — val_recall: 0.737000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m189s\u001B[0m 7s/step - categorical_accuracy: 0.7299 - loss: 0.5328 - val_categorical_accuracy: 0.7370 - val_loss: 0.5390 - val_f1: 0.7367 - val_recall: 0.7370 - val_precision: 0.7370\n",
      "Epoch 4/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m13s\u001B[0m 404ms/step\n",
      " — val_f1: 0.736736 — val_precision: 0.741630 — val_recall: 0.737000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m177s\u001B[0m 6s/step - categorical_accuracy: 0.7243 - loss: 0.5311 - val_categorical_accuracy: 0.7370 - val_loss: 0.5334 - val_f1: 0.7367 - val_recall: 0.7370 - val_precision: 0.7416\n",
      "Epoch 5/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m13s\u001B[0m 414ms/step\n",
      " — val_f1: 0.732105 — val_precision: 0.733358 — val_recall: 0.732000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m195s\u001B[0m 7s/step - categorical_accuracy: 0.7369 - loss: 0.5210 - val_categorical_accuracy: 0.7320 - val_loss: 0.5349 - val_f1: 0.7321 - val_recall: 0.7320 - val_precision: 0.7334\n",
      "Epoch 6/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m12s\u001B[0m 367ms/step\n",
      " — val_f1: 0.743977 — val_precision: 0.746842 — val_recall: 0.744000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m209s\u001B[0m 7s/step - categorical_accuracy: 0.7427 - loss: 0.5159 - val_categorical_accuracy: 0.7440 - val_loss: 0.5329 - val_f1: 0.7440 - val_recall: 0.7440 - val_precision: 0.7468\n",
      "Epoch 7/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m12s\u001B[0m 367ms/step\n",
      " — val_f1: 0.739048 — val_precision: 0.741122 — val_recall: 0.739000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m178s\u001B[0m 6s/step - categorical_accuracy: 0.7455 - loss: 0.5046 - val_categorical_accuracy: 0.7390 - val_loss: 0.5341 - val_f1: 0.7390 - val_recall: 0.7390 - val_precision: 0.7411\n",
      "Epoch 8/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m9s\u001B[0m 292ms/step\n",
      " — val_f1: 0.748109 — val_precision: 0.749079 — val_recall: 0.748000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m188s\u001B[0m 7s/step - categorical_accuracy: 0.7474 - loss: 0.5084 - val_categorical_accuracy: 0.7480 - val_loss: 0.5301 - val_f1: 0.7481 - val_recall: 0.7480 - val_precision: 0.7491\n",
      "Epoch 9/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m10s\u001B[0m 298ms/step\n",
      " — val_f1: 0.736716 — val_precision: 0.748709 — val_recall: 0.738000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m139s\u001B[0m 5s/step - categorical_accuracy: 0.7651 - loss: 0.4742 - val_categorical_accuracy: 0.7380 - val_loss: 0.5396 - val_f1: 0.7367 - val_recall: 0.7380 - val_precision: 0.7487\n",
      "Epoch 10/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m10s\u001B[0m 303ms/step\n",
      " — val_f1: 0.740114 — val_precision: 0.740945 — val_recall: 0.740000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m139s\u001B[0m 5s/step - categorical_accuracy: 0.7524 - loss: 0.4914 - val_categorical_accuracy: 0.7400 - val_loss: 0.5333 - val_f1: 0.7401 - val_recall: 0.7400 - val_precision: 0.7409\n",
      "Epoch 11/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m13s\u001B[0m 391ms/step\n",
      " — val_f1: 0.738084 — val_precision: 0.739680 — val_recall: 0.738000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m181s\u001B[0m 7s/step - categorical_accuracy: 0.7679 - loss: 0.4747 - val_categorical_accuracy: 0.7380 - val_loss: 0.5451 - val_f1: 0.7381 - val_recall: 0.7380 - val_precision: 0.7397\n",
      "Epoch 12/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m12s\u001B[0m 375ms/step\n",
      " — val_f1: 0.734773 — val_precision: 0.739345 — val_recall: 0.735000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m159s\u001B[0m 6s/step - categorical_accuracy: 0.7687 - loss: 0.4711 - val_categorical_accuracy: 0.7350 - val_loss: 0.5469 - val_f1: 0.7348 - val_recall: 0.7350 - val_precision: 0.7393\n",
      "Epoch 13/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m12s\u001B[0m 372ms/step\n",
      " — val_f1: 0.740071 — val_precision: 0.740236 — val_recall: 0.740000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m180s\u001B[0m 6s/step - categorical_accuracy: 0.7653 - loss: 0.4658 - val_categorical_accuracy: 0.7400 - val_loss: 0.5366 - val_f1: 0.7401 - val_recall: 0.7400 - val_precision: 0.7402\n",
      "Epoch 14/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m11s\u001B[0m 340ms/step\n",
      " — val_f1: 0.740000 — val_precision: 0.742615 — val_recall: 0.740000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m230s\u001B[0m 8s/step - categorical_accuracy: 0.7708 - loss: 0.4569 - val_categorical_accuracy: 0.7400 - val_loss: 0.5338 - val_f1: 0.7400 - val_recall: 0.7400 - val_precision: 0.7426\n",
      "Epoch 15/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m14s\u001B[0m 425ms/step\n",
      " — val_f1: 0.730083 — val_precision: 0.739370 — val_recall: 0.731000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m198s\u001B[0m 7s/step - categorical_accuracy: 0.7864 - loss: 0.4446 - val_categorical_accuracy: 0.7310 - val_loss: 0.5563 - val_f1: 0.7301 - val_recall: 0.7310 - val_precision: 0.7394\n",
      "Epoch 16/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m9s\u001B[0m 285ms/step\n",
      " — val_f1: 0.724963 — val_precision: 0.724939 — val_recall: 0.725000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m134s\u001B[0m 5s/step - categorical_accuracy: 0.7926 - loss: 0.4381 - val_categorical_accuracy: 0.7250 - val_loss: 0.5654 - val_f1: 0.7250 - val_recall: 0.7250 - val_precision: 0.7249\n",
      "Epoch 17/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m12s\u001B[0m 382ms/step\n",
      " — val_f1: 0.723976 — val_precision: 0.726761 — val_recall: 0.724000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m141s\u001B[0m 5s/step - categorical_accuracy: 0.7900 - loss: 0.4333 - val_categorical_accuracy: 0.7240 - val_loss: 0.5544 - val_f1: 0.7240 - val_recall: 0.7240 - val_precision: 0.7268\n",
      "Epoch 18/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m9s\u001B[0m 291ms/step\n",
      " — val_f1: 0.728728 — val_precision: 0.733558 — val_recall: 0.729000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m150s\u001B[0m 5s/step - categorical_accuracy: 0.7967 - loss: 0.4213 - val_categorical_accuracy: 0.7290 - val_loss: 0.5643 - val_f1: 0.7287 - val_recall: 0.7290 - val_precision: 0.7336\n",
      "Epoch 19/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m10s\u001B[0m 310ms/step\n",
      " — val_f1: 0.719819 — val_precision: 0.723853 — val_recall: 0.720000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m138s\u001B[0m 5s/step - categorical_accuracy: 0.8061 - loss: 0.4028 - val_categorical_accuracy: 0.7200 - val_loss: 0.5606 - val_f1: 0.7198 - val_recall: 0.7200 - val_precision: 0.7239\n",
      "Epoch 20/20\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m14s\u001B[0m 442ms/step\n",
      " — val_f1: 0.720110 — val_precision: 0.720502 — val_recall: 0.720000\n",
      "\u001B[1m28/28\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m148s\u001B[0m 5s/step - categorical_accuracy: 0.8088 - loss: 0.4020 - val_categorical_accuracy: 0.7200 - val_loss: 0.5929 - val_f1: 0.7201 - val_recall: 0.7200 - val_precision: 0.7205\n"
     ]
    }
   ],
   "source": [
    "# Create an empty sequential model\n",
    "model3 = Sequential()\n",
    "\n",
    "# Add an input layer that accepts a string\n",
    "model3.add(Input(shape=(1,), dtype=tf.string))\n",
    "\n",
    "# Add a vectorizer layer to convert the string into integer tokens\n",
    "model3.add(vectorizer)\n",
    "\n",
    "# Add an embedding layer to convert the integer tokens into dense vectors\n",
    "# The weights of the embedding layer are initialized with a pre-trained embedding (GloVe)\n",
    "model3.add(Embedding(MAX_WORDS, EMBEDDING_DIM, weights=[embedding_matrix],\n",
    "                    input_length=MAX_SEQUENCE_LENGTH, mask_zero=True, trainable=False))\n",
    "\n",
    "# Add a dropout layer to prevent overfitting\n",
    "model3.add(Dropout(0.33))\n",
    "\n",
    "# Add a bidirectional GRU layer\n",
    "model3.add(Bidirectional(GRU(GRU_SIZE, return_sequences=True)))\n",
    "\n",
    "# Add a self-attention layer\n",
    "model3.add(SelfAttention(mlp_layers=2, units=GRU_SIZE*2, dropout_rate=0.33 ))\n",
    "\n",
    "# Add a dense layer with ReLU activation\n",
    "model3.add(Dense(units=DENSE, activation='relu' ))\n",
    "\n",
    "# Add another dropout layer\n",
    "model3.add(Dropout(0.33))\n",
    "\n",
    "# Add the output layer with softmax activation\n",
    "model3.add(Dense(len(y_train_new.unique()), activation='softmax'))\n",
    "\n",
    "# Print the summary of the model\n",
    "print(model3.summary())\n",
    "\n",
    "# Compile the model with categorical crossentropy loss, Adam optimizer and categorical accuracy metric\n",
    "model3.compile(loss='categorical_crossentropy',\n",
    "                  optimizer=Adam(learning_rate=0.001),metrics=[\"categorical_accuracy\"])\n",
    "\n",
    "# Convert the training data into the required format\n",
    "train_data = np.array([[\" \".join(x_train)] for x_train in X_train_tokenized])\n",
    "# Convert the validation data into the required format\n",
    "val_data = np.array([[\" \".join(x_val)] for x_val in X_val_tokenized])\n",
    "\n",
    "# Train the model\n",
    "history3 = model3.fit(train_data.astype(object), y_train_1_hot,\n",
    "              validation_data=(val_data.astype(object), y_val_1_hot),\n",
    "              batch_size=256,\n",
    "              epochs=20,\n",
    "              shuffle=True,\n",
    "              callbacks=[Metrics(valid_data=(val_data.astype(object), y_val_1_hot))])"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:25:43.581117Z",
     "start_time": "2024-09-23T12:30:25.470713Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for accuracy\n",
    "plt.plot(history3.history['categorical_accuracy'])\n",
    "plt.plot(history3.history['val_categorical_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'dev'], loc='upper left')\n",
    "plt.xticks([x_tick for x_tick in range(0,len(history3.history['categorical_accuracy'])+2,2)])\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:26:57.431946Z",
     "start_time": "2024-09-23T13:26:57.295103Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for loss\n",
    "plt.plot(history3.history['loss'])\n",
    "plt.plot(history3.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'dev'], loc='upper right')\n",
    "plt.xticks([x_tick for x_tick in range(0,len(history3.history['categorical_accuracy'])+2,2)])\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:27:04.309039Z",
     "start_time": "2024-09-23T13:27:04.203582Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "outputs": [],
   "source": [
    "# Convert the test data into the required format\n",
    "test_data = np.array([[\" \".join(x_test)] for x_test in X_test_tokenized])"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:27:12.033094Z",
     "start_time": "2024-09-23T13:27:12.019004Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001B[1m219/219\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m50s\u001B[0m 229ms/step\n",
      "\u001B[1m32/32\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m7s\u001B[0m 222ms/step\n",
      "\u001B[1m63/63\u001B[0m \u001B[32m━━━━━━━━━━━━━━━━━━━━\u001B[0m\u001B[37m\u001B[0m \u001B[1m13s\u001B[0m 209ms/step\n"
     ]
    }
   ],
   "source": [
    "# Use the trained RNN model to make predictions on the training data\n",
    "# The model's predict method returns a probability for each class\n",
    "# We use argmax to get the index of the highest probability, which gives us the predicted class\n",
    "predictions_train_rnn = np.argmax(model3.predict(train_data.astype(object)), -1)\n",
    "\n",
    "# Make predictions on the validation data\n",
    "predictions_val_rnn = np.argmax(model3.predict(val_data.astype(object)), -1)\n",
    "\n",
    "# Make predictions on the test data\n",
    "predictions_test_rnn = np.argmax(model3.predict(test_data.astype(object)), -1)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:30:04.485245Z",
     "start_time": "2024-09-23T13:28:53.818646Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report for TRAIN data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.86      0.85      0.86      3499\n",
      "           1       0.85      0.87      0.86      3501\n",
      "\n",
      "    accuracy                           0.86      7000\n",
      "   macro avg       0.86      0.86      0.86      7000\n",
      "weighted avg       0.86      0.86      0.86      7000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Print the classification report for the training data\n",
    "# The classification report shows key classification metrics including precision, recall and f1-score\n",
    "print(\"Classification Report for TRAIN data set:\")\n",
    "print(classification_report(y_train_new, predictions_train_rnn))"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:31:05.610139Z",
     "start_time": "2024-09-23T13:31:05.584552Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report for Validation data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.74      0.72      0.73       521\n",
      "           1       0.70      0.72      0.71       479\n",
      "\n",
      "    accuracy                           0.72      1000\n",
      "   macro avg       0.72      0.72      0.72      1000\n",
      "weighted avg       0.72      0.72      0.72      1000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Print the classification report for the validation data\n",
    "print(\"Classification Report for Validation data set:\")\n",
    "print(classification_report(y_val, predictions_val_rnn))"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:31:07.114483Z",
     "start_time": "2024-09-23T13:31:07.092703Z"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report for TEST data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.73      0.71      0.72       980\n",
      "           1       0.73      0.75      0.74      1020\n",
      "\n",
      "    accuracy                           0.73      2000\n",
      "   macro avg       0.73      0.73      0.73      2000\n",
      "weighted avg       0.73      0.73      0.73      2000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Print the classification report for the test data\n",
    "print(\"Classification Report for TEST data set:\")\n",
    "print(classification_report(y_test, predictions_test_rnn))"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:31:13.200402Z",
     "start_time": "2024-09-23T13:31:13.091749Z"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "#### Hyperparameter tuning for RNN GRU model"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "outputs": [],
   "source": [
    "locale.getpreferredencoding = lambda: \"UTF-8\""
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:31:26.230213Z",
     "start_time": "2024-09-23T13:31:26.207234Z"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "def build_model(hp):\n",
    "    # Initialize a Sequential model\n",
    "    model = Sequential()\n",
    "\n",
    "    # Add an Input layer that accepts a string\n",
    "    model.add(Input(shape=(1,), dtype=tf.string))\n",
    "\n",
    "    # Add a vectorizer layer to convert the string into integer tokens\n",
    "    model.add(vectorizer)\n",
    "\n",
    "    # Add an Embedding layer to convert the integer tokens into dense vectors\n",
    "    # The weights of the Embedding layer are initialized with a pre-trained embedding (GloVe)\n",
    "    model.add(Embedding(MAX_WORDS, EMBEDDING_DIM, weights=[embedding_matrix],\n",
    "                      input_length=MAX_SEQUENCE_LENGTH, mask_zero=True, trainable=False))\n",
    "\n",
    "    # Add a Dropout layer to prevent overfitting\n",
    "    model.add(Dropout(hp.Choice('dropout_rate', values=[0.1,0.2,0.3,0.4,0.5])))\n",
    "\n",
    "    # Add a variable number of Bidirectional GRU layers\n",
    "    for i in range(hp.Int(name='num_rnn_layers',min_value=1,max_value=3)):\n",
    "        model.add(Bidirectional(GRU(hp.Int('gru_units', min_value=128, max_value=512, step=64), return_sequences=True)))\n",
    "\n",
    "    # Add a Self-Attention layer\n",
    "    model.add(SelfAttention(mlp_layers=hp.Int('attention_layers',1,3),\n",
    "                            units=hp.Int('attention_units',min_value=128,max_value=512,step=64),\n",
    "                            dropout_rate=hp.Choice('attention_dropout', values=[0.1,0.2,0.3,0.4,0.5]) ))\n",
    "\n",
    "    # Add a Dense layer with ReLU activation\n",
    "    model.add(Dense(units=hp.Int('dense_units', min_value=128, max_value=512, step=64), activation='relu' ))\n",
    "\n",
    "    # Add another Dropout layer\n",
    "    model.add(Dropout(hp.Choice('dropout_rate_dense', values=[0.1,0.2,0.3,0.4,0.5])))\n",
    "\n",
    "    # Add the output layer with softmax activation\n",
    "    model.add(Dense(len(y_train_new.unique()), activation='softmax'))\n",
    "\n",
    "    # Tune the learning rate for the optimizer\n",
    "    hp_learning_rate = hp.Choice('learning_rate', values=[1e-3, 1e-4])\n",
    "\n",
    "    # Compile the model with categorical crossentropy loss, Adam optimizer and categorical accuracy metric\n",
    "    model.compile(loss='categorical_crossentropy',optimizer=Adam(learning_rate=hp_learning_rate),\n",
    "                  metrics=[CategoricalAccuracy()])\n",
    "\n",
    "    # Return the constructed model\n",
    "    return model"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-09-23T13:31:37.035112Z",
     "start_time": "2024-09-23T13:31:37.010907Z"
    }
   },
   "execution_count": 122,
   "outputs": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "# Remove the directory 'KT_dir' if it exists\n",
    "!rm -r KT_dir"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Search space summary\n",
      "Default search space size: 9\n",
      "dropout_rate (Choice)\n",
      "{'default': 0.1, 'conditions': [], 'values': [0.1, 0.2, 0.3, 0.4, 0.5], 'ordered': True}\n",
      "num_rnn_layers (Int)\n",
      "{'default': None, 'conditions': [], 'min_value': 1, 'max_value': 3, 'step': 1, 'sampling': 'linear'}\n",
      "gru_units (Int)\n",
      "{'default': None, 'conditions': [], 'min_value': 128, 'max_value': 512, 'step': 64, 'sampling': 'linear'}\n",
      "attention_layers (Int)\n",
      "{'default': None, 'conditions': [], 'min_value': 1, 'max_value': 3, 'step': 1, 'sampling': 'linear'}\n",
      "attention_units (Int)\n",
      "{'default': None, 'conditions': [], 'min_value': 128, 'max_value': 512, 'step': 64, 'sampling': 'linear'}\n",
      "attention_dropout (Choice)\n",
      "{'default': 0.1, 'conditions': [], 'values': [0.1, 0.2, 0.3, 0.4, 0.5], 'ordered': True}\n",
      "dense_units (Int)\n",
      "{'default': None, 'conditions': [], 'min_value': 128, 'max_value': 512, 'step': 64, 'sampling': 'linear'}\n",
      "dropout_rate_dense (Choice)\n",
      "{'default': 0.1, 'conditions': [], 'values': [0.1, 0.2, 0.3, 0.4, 0.5], 'ordered': True}\n",
      "learning_rate (Choice)\n",
      "{'default': 0.001, 'conditions': [], 'values': [0.001, 0.0001], 'ordered': True}\n"
     ]
    }
   ],
   "source": [
    "# Initialize a RandomSearch tuner from Keras Tuner\n",
    "# The tuner will search for the best hyperparameters of the model defined in the 'build_model' function\n",
    "# The objective of the tuner is to maximize the validation categorical accuracy\n",
    "# The tuner will perform a maximum of 10 trials\n",
    "# The results of the tuner will be stored in the directory 'KT_dir' under the project name 'KT_tuning'\n",
    "tuner = kt.RandomSearch(build_model,\n",
    "                        objective=kt.Objective(\"val_categorical_accuracy\", direction=\"max\"),\n",
    "                        max_trials = 10,\n",
    "                        directory='KT_dir',\n",
    "                        project_name='KT_tuning')\n",
    "\n",
    "# Initialize an EarlyStopping callback\n",
    "# The callback will stop the training if the validation loss does not improve for 5 epochs\n",
    "early_stopping = EarlyStopping(monitor='val_loss', patience=5)\n",
    "\n",
    "# Print a summary of the search space\n",
    "tuner.search_space_summary()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Trial 10 Complete [00h 13m 30s]\n",
      "val_categorical_accuracy: 0.7429999709129333\n",
      "\n",
      "Best val_categorical_accuracy So Far: 0.7440000176429749\n",
      "Total elapsed time: 02h 23m 13s\n"
     ]
    }
   ],
   "source": [
    "# Start the hyperparameter search\n",
    "# The tuner will train the model on the training data and evaluate it on the validation data\n",
    "# The training will be done for a maximum of 20 epochs with a batch size of 128\n",
    "# The EarlyStopping callback is used during training\n",
    "tuner.search(train_data.astype(object), y_train_1_hot,\n",
    "             validation_data=(val_data, y_val_1_hot), epochs=20, batch_size = 128,\n",
    "             callbacks=[early_stopping])"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Results summary\n",
      "Results in KT_dir/KT_tuning\n",
      "Showing 10 best trials\n",
      "Objective(name=\"val_categorical_accuracy\", direction=\"max\")\n",
      "\n",
      "Trial 01 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.1\n",
      "num_rnn_layers: 1\n",
      "gru_units: 512\n",
      "attention_layers: 3\n",
      "attention_units: 192\n",
      "attention_dropout: 0.4\n",
      "dense_units: 256\n",
      "dropout_rate_dense: 0.2\n",
      "learning_rate: 0.0001\n",
      "Score: 0.7440000176429749\n",
      "\n",
      "Trial 09 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.3\n",
      "num_rnn_layers: 1\n",
      "gru_units: 128\n",
      "attention_layers: 1\n",
      "attention_units: 320\n",
      "attention_dropout: 0.2\n",
      "dense_units: 512\n",
      "dropout_rate_dense: 0.4\n",
      "learning_rate: 0.0001\n",
      "Score: 0.7429999709129333\n",
      "\n",
      "Trial 06 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.4\n",
      "num_rnn_layers: 1\n",
      "gru_units: 384\n",
      "attention_layers: 2\n",
      "attention_units: 448\n",
      "attention_dropout: 0.1\n",
      "dense_units: 320\n",
      "dropout_rate_dense: 0.5\n",
      "learning_rate: 0.001\n",
      "Score: 0.7419999837875366\n",
      "\n",
      "Trial 08 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.4\n",
      "num_rnn_layers: 1\n",
      "gru_units: 320\n",
      "attention_layers: 1\n",
      "attention_units: 128\n",
      "attention_dropout: 0.5\n",
      "dense_units: 512\n",
      "dropout_rate_dense: 0.4\n",
      "learning_rate: 0.001\n",
      "Score: 0.7419999837875366\n",
      "\n",
      "Trial 02 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.2\n",
      "num_rnn_layers: 2\n",
      "gru_units: 512\n",
      "attention_layers: 1\n",
      "attention_units: 512\n",
      "attention_dropout: 0.1\n",
      "dense_units: 384\n",
      "dropout_rate_dense: 0.2\n",
      "learning_rate: 0.0001\n",
      "Score: 0.7400000095367432\n",
      "\n",
      "Trial 07 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.4\n",
      "num_rnn_layers: 1\n",
      "gru_units: 192\n",
      "attention_layers: 2\n",
      "attention_units: 448\n",
      "attention_dropout: 0.5\n",
      "dense_units: 320\n",
      "dropout_rate_dense: 0.3\n",
      "learning_rate: 0.0001\n",
      "Score: 0.7390000224113464\n",
      "\n",
      "Trial 04 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.4\n",
      "num_rnn_layers: 1\n",
      "gru_units: 320\n",
      "attention_layers: 2\n",
      "attention_units: 448\n",
      "attention_dropout: 0.3\n",
      "dense_units: 512\n",
      "dropout_rate_dense: 0.3\n",
      "learning_rate: 0.001\n",
      "Score: 0.7379999756813049\n",
      "\n",
      "Trial 00 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.1\n",
      "num_rnn_layers: 3\n",
      "gru_units: 448\n",
      "attention_layers: 1\n",
      "attention_units: 448\n",
      "attention_dropout: 0.5\n",
      "dense_units: 448\n",
      "dropout_rate_dense: 0.4\n",
      "learning_rate: 0.001\n",
      "Score: 0.7369999885559082\n",
      "\n",
      "Trial 05 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.4\n",
      "num_rnn_layers: 1\n",
      "gru_units: 256\n",
      "attention_layers: 2\n",
      "attention_units: 256\n",
      "attention_dropout: 0.4\n",
      "dense_units: 448\n",
      "dropout_rate_dense: 0.2\n",
      "learning_rate: 0.0001\n",
      "Score: 0.7360000014305115\n",
      "\n",
      "Trial 03 summary\n",
      "Hyperparameters:\n",
      "dropout_rate: 0.2\n",
      "num_rnn_layers: 3\n",
      "gru_units: 192\n",
      "attention_layers: 2\n",
      "attention_units: 512\n",
      "attention_dropout: 0.2\n",
      "dense_units: 320\n",
      "dropout_rate_dense: 0.1\n",
      "learning_rate: 0.0001\n",
      "Score: 0.6909999847412109\n"
     ]
    }
   ],
   "source": [
    "# Print a summary of the results of the hyperparameter search\n",
    "tuner.results_summary()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "{'dropout_rate': 0.1,\n",
       " 'num_rnn_layers': 1,\n",
       " 'gru_units': 512,\n",
       " 'attention_layers': 3,\n",
       " 'attention_units': 192,\n",
       " 'attention_dropout': 0.4,\n",
       " 'dense_units': 256,\n",
       " 'dropout_rate_dense': 0.2,\n",
       " 'learning_rate': 0.0001}"
      ]
     },
     "metadata": {},
     "execution_count": 54
    }
   ],
   "source": [
    "# Get best hyper-parameters setup\n",
    "tuner.get_best_hyperparameters()[0].values"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " text_vectorization (TextVec  (None, 256)              0         \n",
      " torization)                                                     \n",
      "                                                                 \n",
      " embedding (Embedding)       (None, 256, 100)          51200     \n",
      "                                                                 \n",
      " dropout (Dropout)           (None, 256, 100)          0         \n",
      "                                                                 \n",
      " bidirectional (Bidirectiona  (None, 256, 1024)        1886208   \n",
      " l)                                                              \n",
      "                                                                 \n",
      " self_attention (SelfAttenti  (None, 1024)             271105    \n",
      " on)                                                             \n",
      "                                                                 \n",
      " dense_4 (Dense)             (None, 256)               262400    \n",
      "                                                                 \n",
      " dropout_4 (Dropout)         (None, 256)               0         \n",
      "                                                                 \n",
      " dense_5 (Dense)             (None, 2)                 514       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 2,471,427\n",
      "Trainable params: 2,420,227\n",
      "Non-trainable params: 51,200\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Get the best model found during the search\n",
    "best_model_2 = tuner.get_best_models(num_models=1)[0]\n",
    "\n",
    "# Print a summary of the best model\n",
    "best_model_2.summary()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "# Use the best model to predict the class labels for the training data\n",
    "# The model's predict method returns a probability for each class\n",
    "# We use argmax to get the index of the highest probability, which gives us the predicted class\n",
    "predictions_train_rnn = np.argmax(best_model_2.predict(df_tweet_sentiment), -1)\n",
    "\n",
    "# Make predictions on the validation data\n",
    "predictions_val_rnn = np.argmax(best_model_2.predict(val_data), -1)\n",
    "\n",
    "# Make predictions on the test data\n",
    "predictions_test_rnn= np.argmax(best_model_2.predict(test_data), -1)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "219/219 [==============================] - 10s 34ms/step\n",
      "32/32 [==============================] - 1s 33ms/step\n",
      "63/63 [==============================] - 2s 39ms/step\n",
      "Classification Report for TRAIN data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.77      0.79      0.78      3495\n",
      "           1       0.79      0.76      0.77      3505\n",
      "\n",
      "    accuracy                           0.77      7000\n",
      "   macro avg       0.78      0.77      0.77      7000\n",
      "weighted avg       0.78      0.77      0.77      7000\n",
      "\n",
      "Classification Report for Validation data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.76      0.74      0.75       522\n",
      "           1       0.73      0.74      0.74       478\n",
      "\n",
      "    accuracy                           0.74      1000\n",
      "   macro avg       0.74      0.74      0.74      1000\n",
      "weighted avg       0.74      0.74      0.74      1000\n",
      "\n",
      "Classification Report for TEST data set:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.73      0.75      0.74       983\n",
      "           1       0.75      0.73      0.74      1017\n",
      "\n",
      "    accuracy                           0.74      2000\n",
      "   macro avg       0.74      0.74      0.74      2000\n",
      "weighted avg       0.74      0.74      0.74      2000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Print a classification report for the training data\n",
    "# The classification report shows key classification metrics including precision, recall and f1-score\n",
    "print(\"Classification Report for TRAIN data set:\")\n",
    "print(classification_report(y_train_new, predictions_train_rnn))\n",
    "\n",
    "# Print a classification report for the validation data\n",
    "print(\"Classification Report for Validation data set:\")\n",
    "print(classification_report(y_val, predictions_val_rnn))\n",
    "\n",
    "# Print a classification report for the test data\n",
    "print(\"Classification Report for TEST data set:\")\n",
    "print(classification_report(y_test, predictions_test_rnn))"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 1/20\n",
      "32/32 [==============================] - 1s 25ms/step\n",
      " — val_f1: 0.731053 — val_precision: 0.731142 — val_recall: 0.731000\n",
      "28/28 [==============================] - 43s 1s/step - loss: 0.4818 - categorical_accuracy: 0.7650 - val_loss: 0.5299 - val_categorical_accuracy: 0.7320 - val_f1: 0.7311 - val_recall: 0.7310 - val_precision: 0.7311\n",
      "Epoch 2/20\n",
      "32/32 [==============================] - 1s 26ms/step\n",
      " — val_f1: 0.734874 — val_precision: 0.738891 — val_recall: 0.735000\n",
      "28/28 [==============================] - 27s 963ms/step - loss: 0.4791 - categorical_accuracy: 0.7694 - val_loss: 0.5354 - val_categorical_accuracy: 0.7350 - val_f1: 0.7349 - val_recall: 0.7350 - val_precision: 0.7389\n",
      "Epoch 3/20\n",
      "32/32 [==============================] - 1s 24ms/step\n",
      " — val_f1: 0.732091 — val_precision: 0.732329 — val_recall: 0.732000\n",
      "28/28 [==============================] - 34s 1s/step - loss: 0.4798 - categorical_accuracy: 0.7667 - val_loss: 0.5284 - val_categorical_accuracy: 0.7320 - val_f1: 0.7321 - val_recall: 0.7320 - val_precision: 0.7323\n",
      "Epoch 4/20\n",
      "32/32 [==============================] - 1s 25ms/step\n",
      " — val_f1: 0.735107 — val_precision: 0.735462 — val_recall: 0.735000\n",
      "28/28 [==============================] - 29s 1s/step - loss: 0.4755 - categorical_accuracy: 0.7670 - val_loss: 0.5281 - val_categorical_accuracy: 0.7350 - val_f1: 0.7351 - val_recall: 0.7350 - val_precision: 0.7355\n",
      "Epoch 5/20\n",
      "32/32 [==============================] - 1s 24ms/step\n",
      " — val_f1: 0.732905 — val_precision: 0.736635 — val_recall: 0.733000\n",
      "28/28 [==============================] - 31s 1s/step - loss: 0.4726 - categorical_accuracy: 0.7679 - val_loss: 0.5350 - val_categorical_accuracy: 0.7330 - val_f1: 0.7329 - val_recall: 0.7330 - val_precision: 0.7366\n",
      "Epoch 6/20\n",
      "32/32 [==============================] - 1s 25ms/step\n",
      " — val_f1: 0.739124 — val_precision: 0.740172 — val_recall: 0.739000\n",
      "28/28 [==============================] - 30s 1s/step - loss: 0.4715 - categorical_accuracy: 0.7694 - val_loss: 0.5286 - val_categorical_accuracy: 0.7390 - val_f1: 0.7391 - val_recall: 0.7390 - val_precision: 0.7402\n",
      "Epoch 7/20\n",
      "32/32 [==============================] - 1s 25ms/step\n",
      " — val_f1: 0.744098 — val_precision: 0.744411 — val_recall: 0.744000\n",
      "28/28 [==============================] - 33s 1s/step - loss: 0.4739 - categorical_accuracy: 0.7634 - val_loss: 0.5278 - val_categorical_accuracy: 0.7440 - val_f1: 0.7441 - val_recall: 0.7440 - val_precision: 0.7444\n",
      "Epoch 8/20\n",
      "32/32 [==============================] - 1s 31ms/step\n",
      " — val_f1: 0.736127 — val_precision: 0.737101 — val_recall: 0.736000\n",
      "28/28 [==============================] - 27s 947ms/step - loss: 0.4712 - categorical_accuracy: 0.7676 - val_loss: 0.5295 - val_categorical_accuracy: 0.7360 - val_f1: 0.7361 - val_recall: 0.7360 - val_precision: 0.7371\n",
      "Epoch 9/20\n",
      "32/32 [==============================] - 1s 26ms/step\n",
      " — val_f1: 0.737106 — val_precision: 0.738629 — val_recall: 0.737000\n",
      "28/28 [==============================] - 32s 1s/step - loss: 0.4633 - categorical_accuracy: 0.7733 - val_loss: 0.5363 - val_categorical_accuracy: 0.7320 - val_f1: 0.7371 - val_recall: 0.7370 - val_precision: 0.7386\n",
      "Epoch 10/20\n",
      "32/32 [==============================] - 1s 29ms/step\n",
      " — val_f1: 0.736124 — val_precision: 0.737242 — val_recall: 0.736000\n",
      "28/28 [==============================] - 29s 999ms/step - loss: 0.4615 - categorical_accuracy: 0.7727 - val_loss: 0.5336 - val_categorical_accuracy: 0.7360 - val_f1: 0.7361 - val_recall: 0.7360 - val_precision: 0.7372\n",
      "Epoch 11/20\n",
      "32/32 [==============================] - 1s 25ms/step\n",
      " — val_f1: 0.732023 — val_precision: 0.734632 — val_recall: 0.732000\n",
      "28/28 [==============================] - 32s 1s/step - loss: 0.4602 - categorical_accuracy: 0.7769 - val_loss: 0.5374 - val_categorical_accuracy: 0.7320 - val_f1: 0.7320 - val_recall: 0.7320 - val_precision: 0.7346\n",
      "Epoch 12/20\n",
      "32/32 [==============================] - 1s 32ms/step\n",
      " — val_f1: 0.734077 — val_precision: 0.736057 — val_recall: 0.734000\n",
      "28/28 [==============================] - 29s 1s/step - loss: 0.4598 - categorical_accuracy: 0.7774 - val_loss: 0.5374 - val_categorical_accuracy: 0.7340 - val_f1: 0.7341 - val_recall: 0.7340 - val_precision: 0.7361\n"
     ]
    }
   ],
   "source": [
    "# Train the best model on the training data and evaluate it on the validation data\n",
    "# The training will be done for a maximum of 20 epochs with a batch size of 256\n",
    "# The EarlyStopping callback is used during training to stop training if the validation loss doesn't improve for a certain number of epochs\n",
    "history_val_2 = best_model_2.fit(df_tweet_sentiment, y_train_1_hot,\n",
    "                                 validation_data=(val_data, y_val_1_hot),\n",
    "                                 batch_size=256,\n",
    "                                 epochs=20,\n",
    "                                 shuffle=True,\n",
    "                                 callbacks=[Metrics(valid_data=(val_data, y_val_1_hot)),\n",
    "                                            early_stopping])"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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\n"
     },
     "metadata": {}
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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\n"
     },
     "metadata": {}
    }
   ],
   "source": [
    "# summarize history for accuracy\n",
    "plt.plot(history_val_2.history['categorical_accuracy'])\n",
    "plt.plot(history_val_2.history['val_categorical_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'dev'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# summarize history for loss\n",
    "plt.plot(history_val_2.history['loss'])\n",
    "plt.plot(history_val_2.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'dev'], loc='upper right')\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Precision-Recall curves"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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\n"
     },
     "metadata": {}
    }
   ],
   "source": [
    "#disable warnings for large float numbers of estimated probabilities\n",
    "np.seterr(all='ignore')\n",
    "name = 'TunedLogistic Regression Tf-Idf'\n",
    "model = log_cv\n",
    "model.fit(x_train_tfidf, y_train)\n",
    "pred = model.predict_proba(x_test_tfidf)\n",
    "precision, recall, thresholds = precision_recall_curve(y_test, pred[:,1])\n",
    "area = auc(recall, precision)\n",
    "\n",
    "plt.plot(recall, precision, label='Precision-Recall curve')\n",
    "plt.xlabel('Recall')\n",
    "plt.ylabel('Precision')\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.title('Precision-Recall %s: AUC=%0.2f' % (name,area))\n",
    "plt.legend(loc=\"lower left\")\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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\n"
     },
     "metadata": {}
    }
   ],
   "source": [
    "#disable warnings for large float numbers of estimated probabilities\n",
    "np.seterr(all='ignore')\n",
    "name = 'MLP With Centroid'\n",
    "model = model2\n",
    "pred = predictions\n",
    "precision, recall, thresholds = precision_recall_curve(y_test, pred)\n",
    "area = auc(recall, precision)\n",
    "\n",
    "plt.plot(recall, precision, label='Precision-Recall curve')\n",
    "plt.xlabel('Recall')\n",
    "plt.ylabel('Precision')\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.title('Precision-Recall %s: AUC=%0.2f' % (name,area))\n",
    "plt.legend(loc=\"lower left\")\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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\n"
     },
     "metadata": {}
    }
   ],
   "source": [
    "#disable warnings for large float numbers of estimated probabilities\n",
    "np.seterr(all='ignore')\n",
    "name = 'RNN-GRU HyperTuned'\n",
    "model = best_model_2\n",
    "pred = predictions_test_rnn\n",
    "precision, recall, thresholds = precision_recall_curve(y_test, pred)\n",
    "area = auc(recall, precision)\n",
    "\n",
    "plt.plot(recall, precision, label='Precision-Recall curve')\n",
    "plt.xlabel('Recall')\n",
    "plt.ylabel('Precision')\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.title('Precision-Recall %s: AUC=%0.2f' % (name,area))\n",
    "plt.legend(loc=\"lower left\")\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  }
 ],
 "metadata": {
  "colab": {
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}