Notebook/twitter-sentiment-analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "088d0bda",
   "metadata": {
    "papermill": {
     "duration": 0.055644,
     "end_time": "2022-04-16T08:51:13.002714",
     "exception": false,
     "start_time": "2022-04-16T08:51:12.947070",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Twitter Sentiment Analysis by Ibrahim SEROUIS 💻\n",
    "\n",
    "## What is Sentiment Analysis ? 🔍\n",
    "\n",
    "Sentiment analysis is a natural language processing technique used to identify the emotional tone of pieces of writing. In most cases, goal is to determine whether the corpus is postive, negative or neutral. \n",
    "\n",
    "## What to expect 🤔\n",
    "\n",
    "In this Notebook, we're going to train a Sentiment Analysis model on 1.6 million tweets extracted from the well-known Sentiment140 Dataset. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c8d7fdac",
   "metadata": {
    "id": "9ZXB7WM81ygu",
    "papermill": {
     "duration": 0.054962,
     "end_time": "2022-04-16T08:51:13.111745",
     "exception": false,
     "start_time": "2022-04-16T08:51:13.056783",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e54c7bf5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:13.225851Z",
     "iopub.status.busy": "2022-04-16T08:51:13.224349Z",
     "iopub.status.idle": "2022-04-16T08:51:18.254919Z",
     "shell.execute_reply": "2022-04-16T08:51:18.255356Z",
     "shell.execute_reply.started": "2022-04-16T08:49:18.442441Z"
    },
    "id": "IhWt6DDX1yB8",
    "outputId": "7053acb5-17e2-45c8-fc22-8074d9018976",
    "papermill": {
     "duration": 5.089422,
     "end_time": "2022-04-16T08:51:18.255610",
     "exception": false,
     "start_time": "2022-04-16T08:51:13.166188",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-04-16 08:51:14.656076: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0\n"
     ]
    }
   ],
   "source": [
    "# Regular expressions\n",
    "import re\n",
    "\n",
    "# Data and visualization \n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Calculations\n",
    "import numpy as np\n",
    "\n",
    "# Machine Learning\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "# Deep LEarning\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.layers import Embedding, LSTM\n",
    "from tensorflow.keras.layers.experimental.preprocessing import TextVectorization\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d8241700",
   "metadata": {
    "id": "8-hCGQqW2o2D",
    "papermill": {
     "duration": 0.05354,
     "end_time": "2022-04-16T08:51:18.363424",
     "exception": false,
     "start_time": "2022-04-16T08:51:18.309884",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Exploring the dataset 📊"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3a0fef5e",
   "metadata": {
    "id": "OoY1m43i24nU",
    "papermill": {
     "duration": 0.053735,
     "end_time": "2022-04-16T08:51:18.470993",
     "exception": false,
     "start_time": "2022-04-16T08:51:18.417258",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Load"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "4a1ef7a3",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:18.583798Z",
     "iopub.status.busy": "2022-04-16T08:51:18.583250Z",
     "iopub.status.idle": "2022-04-16T08:51:23.728901Z",
     "shell.execute_reply": "2022-04-16T08:51:23.727909Z",
     "shell.execute_reply.started": "2022-04-16T08:49:23.618609Z"
    },
    "id": "k0bHQUy31HL_",
    "papermill": {
     "duration": 5.204039,
     "end_time": "2022-04-16T08:51:23.729061",
     "exception": false,
     "start_time": "2022-04-16T08:51:18.525022",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "dataset = pd.read_csv(\"/kaggle/input/twitter-sentiment-analysis-and-word-embeddings/Dataset/Dataset/training_dataset.csv\", \n",
    "                      encoding=\"latin\", \n",
    "                      header=None, \n",
    "                      names=[\"target\",\"id\",\"date\",\"query\",\"username\",\"content\"]\n",
    "                     )"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6d5a492d",
   "metadata": {
    "id": "o9_eOnd67Qpr",
    "papermill": {
     "duration": 0.056837,
     "end_time": "2022-04-16T08:51:23.844131",
     "exception": false,
     "start_time": "2022-04-16T08:51:23.787294",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Show some examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "b173807c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:23.962480Z",
     "iopub.status.busy": "2022-04-16T08:51:23.961726Z",
     "iopub.status.idle": "2022-04-16T08:51:23.978117Z",
     "shell.execute_reply": "2022-04-16T08:51:23.978513Z",
     "shell.execute_reply.started": "2022-04-16T08:49:29.134194Z"
    },
    "id": "tPuZ_Aht7QC5",
    "outputId": "b79de459-6d91-411b-feb3-1cb39e67a14a",
    "papermill": {
     "duration": 0.079836,
     "end_time": "2022-04-16T08:51:23.978633",
     "exception": false,
     "start_time": "2022-04-16T08:51:23.898797",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>id</th>\n",
       "      <th>date</th>\n",
       "      <th>query</th>\n",
       "      <th>username</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810369</td>\n",
       "      <td>Mon Apr 06 22:19:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_TheSpecialOne_</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810672</td>\n",
       "      <td>Mon Apr 06 22:19:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>scotthamilton</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810917</td>\n",
       "      <td>Mon Apr 06 22:19:53 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mattycus</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811184</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ElleCTF</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811193</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Karoli</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   target          id                          date     query  \\\n",
       "0       0  1467810369  Mon Apr 06 22:19:45 PDT 2009  NO_QUERY   \n",
       "1       0  1467810672  Mon Apr 06 22:19:49 PDT 2009  NO_QUERY   \n",
       "2       0  1467810917  Mon Apr 06 22:19:53 PDT 2009  NO_QUERY   \n",
       "3       0  1467811184  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "4       0  1467811193  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "\n",
       "          username                                            content  \n",
       "0  _TheSpecialOne_  @switchfoot http://twitpic.com/2y1zl - Awww, t...  \n",
       "1    scotthamilton  is upset that he can't update his Facebook by ...  \n",
       "2         mattycus  @Kenichan I dived many times for the ball. Man...  \n",
       "3          ElleCTF    my whole body feels itchy and like its on fire   \n",
       "4           Karoli  @nationwideclass no, it's not behaving at all....  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "13124432",
   "metadata": {
    "id": "2YaoqG3b9C0Y",
    "papermill": {
     "duration": 0.061093,
     "end_time": "2022-04-16T08:51:24.100109",
     "exception": false,
     "start_time": "2022-04-16T08:51:24.039016",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Check for missing values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ee3fff6c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:24.229378Z",
     "iopub.status.busy": "2022-04-16T08:51:24.228598Z",
     "iopub.status.idle": "2022-04-16T08:51:24.797127Z",
     "shell.execute_reply": "2022-04-16T08:51:24.797541Z",
     "shell.execute_reply.started": "2022-04-16T07:33:11.957135Z"
    },
    "id": "xhED0yWp7ihG",
    "outputId": "69eba5ef-81e0-41ca-9af0-3c2b55c4d0eb",
    "papermill": {
     "duration": 0.629366,
     "end_time": "2022-04-16T08:51:24.797703",
     "exception": false,
     "start_time": "2022-04-16T08:51:24.168337",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Missing values: \n",
      "\n",
      "target      0\n",
      "id          0\n",
      "date        0\n",
      "query       0\n",
      "username    0\n",
      "content     0\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print(f\"Missing values: \\n\\n{dataset.isna().sum()}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9bb6c1f5",
   "metadata": {
    "id": "JO4eJJq-9xcr",
    "papermill": {
     "duration": 0.054356,
     "end_time": "2022-04-16T08:51:24.907366",
     "exception": false,
     "start_time": "2022-04-16T08:51:24.853010",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "\n",
    "\n",
    "```\n",
    " There are 0 missing values within this dataset\n",
    "```\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d552c0b4",
   "metadata": {
    "id": "cSq9QERv-jvp",
    "papermill": {
     "duration": 0.060216,
     "end_time": "2022-04-16T08:51:25.022250",
     "exception": false,
     "start_time": "2022-04-16T08:51:24.962034",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Target distribution "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0d8bdfa8",
   "metadata": {
    "id": "K8tlUnkH-nZQ",
    "papermill": {
     "duration": 0.055079,
     "end_time": "2022-04-16T08:51:25.132673",
     "exception": false,
     "start_time": "2022-04-16T08:51:25.077594",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Negative samples : (*target = 0*) 😒"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "6676ac63",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:25.248336Z",
     "iopub.status.busy": "2022-04-16T08:51:25.247701Z",
     "iopub.status.idle": "2022-04-16T08:51:25.333992Z",
     "shell.execute_reply": "2022-04-16T08:51:25.334502Z",
     "shell.execute_reply.started": "2022-04-16T07:33:12.532224Z"
    },
    "id": "Uv8bxTGD9V3l",
    "outputId": "0c83028f-78f0-4bd5-d8eb-c44dd6a1dc47",
    "papermill": {
     "duration": 0.147322,
     "end_time": "2022-04-16T08:51:25.334697",
     "exception": false,
     "start_time": "2022-04-16T08:51:25.187375",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of negative samples : 800000\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>id</th>\n",
       "      <th>date</th>\n",
       "      <th>query</th>\n",
       "      <th>username</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810369</td>\n",
       "      <td>Mon Apr 06 22:19:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_TheSpecialOne_</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810672</td>\n",
       "      <td>Mon Apr 06 22:19:49 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>scotthamilton</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1467810917</td>\n",
       "      <td>Mon Apr 06 22:19:53 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>mattycus</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811184</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ElleCTF</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1467811193</td>\n",
       "      <td>Mon Apr 06 22:19:57 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Karoli</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   target          id                          date     query  \\\n",
       "0       0  1467810369  Mon Apr 06 22:19:45 PDT 2009  NO_QUERY   \n",
       "1       0  1467810672  Mon Apr 06 22:19:49 PDT 2009  NO_QUERY   \n",
       "2       0  1467810917  Mon Apr 06 22:19:53 PDT 2009  NO_QUERY   \n",
       "3       0  1467811184  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "4       0  1467811193  Mon Apr 06 22:19:57 PDT 2009  NO_QUERY   \n",
       "\n",
       "          username                                            content  \n",
       "0  _TheSpecialOne_  @switchfoot http://twitpic.com/2y1zl - Awww, t...  \n",
       "1    scotthamilton  is upset that he can't update his Facebook by ...  \n",
       "2         mattycus  @Kenichan I dived many times for the ball. Man...  \n",
       "3          ElleCTF    my whole body feels itchy and like its on fire   \n",
       "4           Karoli  @nationwideclass no, it's not behaving at all....  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "negative_samples = dataset[dataset[\"target\"]==0]\n",
    "print(f\"Number of negative samples : {len(negative_samples)}\\n\")\n",
    "negative_samples.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ccca96e",
   "metadata": {
    "id": "KM1ZuqFEA6x3",
    "papermill": {
     "duration": 0.054986,
     "end_time": "2022-04-16T08:51:25.445795",
     "exception": false,
     "start_time": "2022-04-16T08:51:25.390809",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Neutral samples : (*target=2*) 😐"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "86671eac",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:25.566444Z",
     "iopub.status.busy": "2022-04-16T08:51:25.565620Z",
     "iopub.status.idle": "2022-04-16T08:51:25.576147Z",
     "shell.execute_reply": "2022-04-16T08:51:25.576570Z",
     "shell.execute_reply.started": "2022-04-16T07:33:12.623121Z"
    },
    "id": "6DT7jZMCAecG",
    "outputId": "3ada8019-9f04-4135-c142-f622fe23164e",
    "papermill": {
     "duration": 0.075779,
     "end_time": "2022-04-16T08:51:25.576725",
     "exception": false,
     "start_time": "2022-04-16T08:51:25.500946",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of neutral samples : 0\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>id</th>\n",
       "      <th>date</th>\n",
       "      <th>query</th>\n",
       "      <th>username</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [target, id, date, query, username, content]\n",
       "Index: []"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "neutral_samples = dataset[dataset[\"target\"]==2]\n",
    "print(f\"Number of neutral samples : {len(neutral_samples)}\\n\\n\")\n",
    "neutral_samples.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "90a61809",
   "metadata": {
    "id": "VHOQYpNmCL6-",
    "papermill": {
     "duration": 0.058902,
     "end_time": "2022-04-16T08:51:25.697400",
     "exception": false,
     "start_time": "2022-04-16T08:51:25.638498",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "\n",
    "\n",
    "```\n",
    "0 neutral sample ? Wow\n",
    "```\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "137bd81a",
   "metadata": {
    "id": "NqkK7OZXCROU",
    "papermill": {
     "duration": 0.056977,
     "end_time": "2022-04-16T08:51:25.812975",
     "exception": false,
     "start_time": "2022-04-16T08:51:25.755998",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Positive samples (*target=4*) 😄"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "73fcab59",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:25.928605Z",
     "iopub.status.busy": "2022-04-16T08:51:25.927447Z",
     "iopub.status.idle": "2022-04-16T08:51:25.997891Z",
     "shell.execute_reply": "2022-04-16T08:51:25.997403Z",
     "shell.execute_reply.started": "2022-04-16T07:33:12.638295Z"
    },
    "id": "sre7a5oBCQ1Q",
    "outputId": "3b46ce62-a57a-400a-900c-5edfe72418b7",
    "papermill": {
     "duration": 0.129271,
     "end_time": "2022-04-16T08:51:25.998026",
     "exception": false,
     "start_time": "2022-04-16T08:51:25.868755",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of positive samples 800000\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>id</th>\n",
       "      <th>date</th>\n",
       "      <th>query</th>\n",
       "      <th>username</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>800000</th>\n",
       "      <td>4</td>\n",
       "      <td>1467822272</td>\n",
       "      <td>Mon Apr 06 22:22:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>ersle</td>\n",
       "      <td>I LOVE @Health4UandPets u guys r the best!!</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800001</th>\n",
       "      <td>4</td>\n",
       "      <td>1467822273</td>\n",
       "      <td>Mon Apr 06 22:22:45 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>becca210</td>\n",
       "      <td>im meeting up with one of my besties tonight! ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800002</th>\n",
       "      <td>4</td>\n",
       "      <td>1467822283</td>\n",
       "      <td>Mon Apr 06 22:22:46 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>Wingman29</td>\n",
       "      <td>@DaRealSunisaKim Thanks for the Twitter add, S...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800003</th>\n",
       "      <td>4</td>\n",
       "      <td>1467822287</td>\n",
       "      <td>Mon Apr 06 22:22:46 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>katarinka</td>\n",
       "      <td>Being sick can be really cheap when it hurts t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800004</th>\n",
       "      <td>4</td>\n",
       "      <td>1467822293</td>\n",
       "      <td>Mon Apr 06 22:22:46 PDT 2009</td>\n",
       "      <td>NO_QUERY</td>\n",
       "      <td>_EmilyYoung</td>\n",
       "      <td>@LovesBrooklyn2 he has that effect on everyone</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        target          id                          date     query  \\\n",
       "800000       4  1467822272  Mon Apr 06 22:22:45 PDT 2009  NO_QUERY   \n",
       "800001       4  1467822273  Mon Apr 06 22:22:45 PDT 2009  NO_QUERY   \n",
       "800002       4  1467822283  Mon Apr 06 22:22:46 PDT 2009  NO_QUERY   \n",
       "800003       4  1467822287  Mon Apr 06 22:22:46 PDT 2009  NO_QUERY   \n",
       "800004       4  1467822293  Mon Apr 06 22:22:46 PDT 2009  NO_QUERY   \n",
       "\n",
       "           username                                            content  \n",
       "800000        ersle       I LOVE @Health4UandPets u guys r the best!!   \n",
       "800001     becca210  im meeting up with one of my besties tonight! ...  \n",
       "800002    Wingman29  @DaRealSunisaKim Thanks for the Twitter add, S...  \n",
       "800003    katarinka  Being sick can be really cheap when it hurts t...  \n",
       "800004  _EmilyYoung    @LovesBrooklyn2 he has that effect on everyone   "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "positive_samples = dataset[dataset[\"target\"]==4]\n",
    "print(f\"Number of positive samples {len(positive_samples)}\\n\\n\")\n",
    "positive_samples.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "74a93fe0",
   "metadata": {
    "id": "tZYiR8QDDfnp",
    "papermill": {
     "duration": 0.056948,
     "end_time": "2022-04-16T08:51:26.112891",
     "exception": false,
     "start_time": "2022-04-16T08:51:26.055943",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Distribution plot"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8d91961a",
   "metadata": {
    "id": "7Am3AkLNEF9e",
    "papermill": {
     "duration": 0.056165,
     "end_time": "2022-04-16T08:51:26.228514",
     "exception": false,
     "start_time": "2022-04-16T08:51:26.172349",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### Defining data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "3fad61c2",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:26.346610Z",
     "iopub.status.busy": "2022-04-16T08:51:26.345913Z",
     "iopub.status.idle": "2022-04-16T08:51:26.348092Z",
     "shell.execute_reply": "2022-04-16T08:51:26.348454Z",
     "shell.execute_reply.started": "2022-04-16T07:33:12.710068Z"
    },
    "id": "5EdHWdBdDej7",
    "papermill": {
     "duration": 0.062383,
     "end_time": "2022-04-16T08:51:26.348579",
     "exception": false,
     "start_time": "2022-04-16T08:51:26.286196",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "data = [len(negative_samples), len(neutral_samples), len(positive_samples)]\n",
    "labels = [\"Negative\",\"Neutral\",\"Positive\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d76603c5",
   "metadata": {
    "id": "Uj1-OsBnEJrV",
    "papermill": {
     "duration": 0.05599,
     "end_time": "2022-04-16T08:51:26.460762",
     "exception": false,
     "start_time": "2022-04-16T08:51:26.404772",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### Define color palette : See [seaborn color palettes](https://seaborn.pydata.org/tutorial/color_palettes.html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "416c8dc6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:26.578620Z",
     "iopub.status.busy": "2022-04-16T08:51:26.577139Z",
     "iopub.status.idle": "2022-04-16T08:51:26.579340Z",
     "shell.execute_reply": "2022-04-16T08:51:26.579760Z",
     "shell.execute_reply.started": "2022-04-16T07:33:12.717141Z"
    },
    "id": "pXz_gP0-DdNQ",
    "papermill": {
     "duration": 0.062797,
     "end_time": "2022-04-16T08:51:26.579885",
     "exception": false,
     "start_time": "2022-04-16T08:51:26.517088",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "colors = sns.color_palette(\"pastel\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9052eb69",
   "metadata": {
    "id": "ESqP10mDEUhf",
    "papermill": {
     "duration": 0.056261,
     "end_time": "2022-04-16T08:51:26.692417",
     "exception": false,
     "start_time": "2022-04-16T08:51:26.636156",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### Dsitribution pie chart"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "daa0d364",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:26.824523Z",
     "iopub.status.busy": "2022-04-16T08:51:26.823895Z",
     "iopub.status.idle": "2022-04-16T08:51:26.912207Z",
     "shell.execute_reply": "2022-04-16T08:51:26.912586Z",
     "shell.execute_reply.started": "2022-04-16T07:33:12.72668Z"
    },
    "id": "YGGg2baaETLr",
    "outputId": "ed0e2df0-c414-42ce-e69c-6e4320857c1c",
    "papermill": {
     "duration": 0.164192,
     "end_time": "2022-04-16T08:51:26.912732",
     "exception": false,
     "start_time": "2022-04-16T08:51:26.748540",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(6,6))\n",
    "plt.title(\"Target distribution\")\n",
    "plt.pie(data, labels=labels, colors=colors, autopct=\"%.0f%%\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4544cef0",
   "metadata": {
    "id": "RTZEZqv_HZOE",
    "papermill": {
     "duration": 0.056608,
     "end_time": "2022-04-16T08:51:27.026408",
     "exception": false,
     "start_time": "2022-04-16T08:51:26.969800",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Data cleaning 🧽🧼"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf011c09",
   "metadata": {
    "id": "yz1BulVbHdGl",
    "papermill": {
     "duration": 0.092069,
     "end_time": "2022-04-16T08:51:27.182478",
     "exception": false,
     "start_time": "2022-04-16T08:51:27.090409",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Drop unncessary columns\n",
    "These columns are not necessary to train our model : ID, date, query, username. \n",
    "\n",
    "> Notes : \n",
    "- axis = 1 means \"computing along the columns\"\n",
    "- using (inplace = true) will overwrite the existing dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "003e7d87",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:27.397717Z",
     "iopub.status.busy": "2022-04-16T08:51:27.394333Z",
     "iopub.status.idle": "2022-04-16T08:51:27.446014Z",
     "shell.execute_reply": "2022-04-16T08:51:27.446793Z",
     "shell.execute_reply.started": "2022-04-16T07:35:17.398398Z"
    },
    "id": "WyZ8e3VLEjfL",
    "papermill": {
     "duration": 0.149463,
     "end_time": "2022-04-16T08:51:27.446946",
     "exception": false,
     "start_time": "2022-04-16T08:51:27.297483",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "dataset.drop([\"id\",\"date\",\"query\",\"username\"],axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1d3f6c43",
   "metadata": {
    "id": "UMcitFGUHzPC",
    "papermill": {
     "duration": 0.180464,
     "end_time": "2022-04-16T08:51:27.721647",
     "exception": false,
     "start_time": "2022-04-16T08:51:27.541183",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "c364d34a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:27.992386Z",
     "iopub.status.busy": "2022-04-16T08:51:27.991581Z",
     "iopub.status.idle": "2022-04-16T08:51:27.998236Z",
     "shell.execute_reply": "2022-04-16T08:51:27.998874Z",
     "shell.execute_reply.started": "2022-04-16T05:04:30.574165Z"
    },
    "id": "KFdq0LhqHysm",
    "outputId": "9a7a159c-7c79-4bef-b162-f07fd3df19d4",
    "papermill": {
     "duration": 0.123274,
     "end_time": "2022-04-16T08:51:27.999063",
     "exception": false,
     "start_time": "2022-04-16T08:51:27.875789",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>@switchfoot http://twitpic.com/2y1zl - Awww, t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>is upset that he can't update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>@Kenichan I dived many times for the ball. Man...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>@nationwideclass no, it's not behaving at all....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   target                                            content\n",
       "0       0  @switchfoot http://twitpic.com/2y1zl - Awww, t...\n",
       "1       0  is upset that he can't update his Facebook by ...\n",
       "2       0  @Kenichan I dived many times for the ball. Man...\n",
       "3       0    my whole body feels itchy and like its on fire \n",
       "4       0  @nationwideclass no, it's not behaving at all...."
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "79d4921b",
   "metadata": {
    "papermill": {
     "duration": 0.095852,
     "end_time": "2022-04-16T08:51:28.189852",
     "exception": false,
     "start_time": "2022-04-16T08:51:28.094000",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Replace the targets \n",
    "\n",
    "In the original dataset, 4 means \"true\". For convenience purposes, we'll set it to 1. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "5b744a5a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:28.390234Z",
     "iopub.status.busy": "2022-04-16T08:51:28.389041Z",
     "iopub.status.idle": "2022-04-16T08:51:28.405625Z",
     "shell.execute_reply": "2022-04-16T08:51:28.406366Z",
     "shell.execute_reply.started": "2022-04-16T05:04:31.736911Z"
    },
    "papermill": {
     "duration": 0.121464,
     "end_time": "2022-04-16T08:51:28.406580",
     "exception": false,
     "start_time": "2022-04-16T08:51:28.285116",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "dataset.target = dataset.target.replace({4: 1})"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "30cc93ca",
   "metadata": {
    "papermill": {
     "duration": 0.094631,
     "end_time": "2022-04-16T08:51:28.596895",
     "exception": false,
     "start_time": "2022-04-16T08:51:28.502264",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "a51412db",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:28.796264Z",
     "iopub.status.busy": "2022-04-16T08:51:28.795520Z",
     "iopub.status.idle": "2022-04-16T08:51:28.852816Z",
     "shell.execute_reply": "2022-04-16T08:51:28.853303Z",
     "shell.execute_reply.started": "2022-04-16T05:04:33.418689Z"
    },
    "papermill": {
     "duration": 0.160789,
     "end_time": "2022-04-16T08:51:28.853467",
     "exception": false,
     "start_time": "2022-04-16T08:51:28.692678",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>800000</th>\n",
       "      <td>1</td>\n",
       "      <td>I LOVE @Health4UandPets u guys r the best!!</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800001</th>\n",
       "      <td>1</td>\n",
       "      <td>im meeting up with one of my besties tonight! ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800002</th>\n",
       "      <td>1</td>\n",
       "      <td>@DaRealSunisaKim Thanks for the Twitter add, S...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800003</th>\n",
       "      <td>1</td>\n",
       "      <td>Being sick can be really cheap when it hurts t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800004</th>\n",
       "      <td>1</td>\n",
       "      <td>@LovesBrooklyn2 he has that effect on everyone</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        target                                            content\n",
       "800000       1       I LOVE @Health4UandPets u guys r the best!! \n",
       "800001       1  im meeting up with one of my besties tonight! ...\n",
       "800002       1  @DaRealSunisaKim Thanks for the Twitter add, S...\n",
       "800003       1  Being sick can be really cheap when it hurts t...\n",
       "800004       1    @LovesBrooklyn2 he has that effect on everyone "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset[dataset[\"target\"]==1].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ac028719",
   "metadata": {
    "id": "UBAfnoUyISya",
    "papermill": {
     "duration": 0.057658,
     "end_time": "2022-04-16T08:51:28.969812",
     "exception": false,
     "start_time": "2022-04-16T08:51:28.912154",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Removing mentions, links, extra spaces from tweets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "adfd436a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:29.094843Z",
     "iopub.status.busy": "2022-04-16T08:51:29.093211Z",
     "iopub.status.idle": "2022-04-16T08:51:29.095508Z",
     "shell.execute_reply": "2022-04-16T08:51:29.095963Z",
     "shell.execute_reply.started": "2022-04-16T05:04:35.630171Z"
    },
    "id": "ikqW_T2zH13p",
    "papermill": {
     "duration": 0.068482,
     "end_time": "2022-04-16T08:51:29.096109",
     "exception": false,
     "start_time": "2022-04-16T08:51:29.027627",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Remove mentions\n",
    "regex_mentions = r\"@[A-Za-z0-9_]+\"\n",
    "# Remove links\n",
    "regex_links = r\"https?://[A-Za-z0-9./]+\"\n",
    "# Remove some special characters\n",
    "regex_special = r\"[^A-Za-z0-9]+\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cd86f91a",
   "metadata": {
    "id": "I9jaNykeM_yj",
    "papermill": {
     "duration": 0.062151,
     "end_time": "2022-04-16T08:51:29.221442",
     "exception": false,
     "start_time": "2022-04-16T08:51:29.159291",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Apply preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "3aaafb7f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:29.383987Z",
     "iopub.status.busy": "2022-04-16T08:51:29.383053Z",
     "iopub.status.idle": "2022-04-16T08:51:47.195019Z",
     "shell.execute_reply": "2022-04-16T08:51:47.194497Z",
     "shell.execute_reply.started": "2022-04-16T05:04:39.882395Z"
    },
    "id": "EoABeZwEM-r1",
    "papermill": {
     "duration": 17.912132,
     "end_time": "2022-04-16T08:51:47.195157",
     "exception": false,
     "start_time": "2022-04-16T08:51:29.283025",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Remove mentions\n",
    "dataset.content = dataset.content.apply(lambda x: re.sub(regex_mentions, \" \", str(x).strip()))\n",
    "# Remove links \n",
    "dataset.content = dataset.content.apply(lambda x: re.sub(regex_links, \" \", str(x).strip()))\n",
    "# Remove special characters\n",
    "dataset.content = dataset.content.apply(lambda x: re.sub(regex_special, \" \", str(x).strip()))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "01bf7a48",
   "metadata": {
    "id": "FqB9ujzoNZ16",
    "papermill": {
     "duration": 0.060992,
     "end_time": "2022-04-16T08:51:47.314875",
     "exception": false,
     "start_time": "2022-04-16T08:51:47.253883",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Show results "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "0e1f4771",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:47.448008Z",
     "iopub.status.busy": "2022-04-16T08:51:47.447158Z",
     "iopub.status.idle": "2022-04-16T08:51:47.450884Z",
     "shell.execute_reply": "2022-04-16T08:51:47.450389Z",
     "shell.execute_reply.started": "2022-04-16T05:04:57.908369Z"
    },
    "id": "0IwBkikKNG-6",
    "papermill": {
     "duration": 0.073995,
     "end_time": "2022-04-16T08:51:47.450999",
     "exception": false,
     "start_time": "2022-04-16T08:51:47.377004",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>target</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>Awww that s a bummer You shoulda got David Ca...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>is upset that he can t update his Facebook by ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>I dived many times for the ball Managed to sav...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>my whole body feels itchy and like its on fire</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>no it s not behaving at all i m mad why am i h...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>not the whole crew</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>Need a hug</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>hey long time no see Yes Rains a bit only a bi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>nope they didn t have it</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0</td>\n",
       "      <td>que me muera</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   target                                            content\n",
       "0       0   Awww that s a bummer You shoulda got David Ca...\n",
       "1       0  is upset that he can t update his Facebook by ...\n",
       "2       0  I dived many times for the ball Managed to sav...\n",
       "3       0     my whole body feels itchy and like its on fire\n",
       "4       0  no it s not behaving at all i m mad why am i h...\n",
       "5       0                                 not the whole crew\n",
       "6       0                                         Need a hug\n",
       "7       0  hey long time no see Yes Rains a bit only a bi...\n",
       "8       0                           nope they didn t have it\n",
       "9       0                                      que me muera "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "191751fe",
   "metadata": {
    "papermill": {
     "duration": 0.060783,
     "end_time": "2022-04-16T08:51:47.573892",
     "exception": false,
     "start_time": "2022-04-16T08:51:47.513109",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Check for null content after cleaning\n",
    "In some rare cases, tweets may contain only special characters and end up \"empty\" after cleaning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "e24782a8",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:47.864306Z",
     "iopub.status.busy": "2022-04-16T08:51:47.863458Z",
     "iopub.status.idle": "2022-04-16T08:51:47.872692Z",
     "shell.execute_reply": "2022-04-16T08:51:47.873321Z",
     "shell.execute_reply.started": "2022-04-16T05:04:57.921066Z"
    },
    "papermill": {
     "duration": 0.239155,
     "end_time": "2022-04-16T08:51:47.873519",
     "exception": false,
     "start_time": "2022-04-16T08:51:47.634364",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Null values: \n",
      "\n",
      "target     0\n",
      "content    0\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print(f\"Null values: \\n\\n{dataset.isna().sum()}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "245349c8",
   "metadata": {
    "id": "4v0DrR2mOwG-",
    "papermill": {
     "duration": 0.064786,
     "end_time": "2022-04-16T08:51:48.004673",
     "exception": false,
     "start_time": "2022-04-16T08:51:47.939887",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Data preparation 👨🏾‍🔬"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7e195ed2",
   "metadata": {
    "papermill": {
     "duration": 0.069461,
     "end_time": "2022-04-16T08:51:48.175695",
     "exception": false,
     "start_time": "2022-04-16T08:51:48.106234",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Train/test split\n",
    "\n",
    "Since the dataset is large enough, we will proceed to a 90-10 (90% for training and 10% for validation/test) split. \n",
    "\n",
    "> Note : Fixing the random state parameter makes our splitting operation reproductible by initializing the internal random number generator which will decide the splitting of data into train and test indices in our case. \n",
    "\n",
    "Inspired by [this answer](https://intellipaat.com/community/15379/python-random-state-in-splitting-dataset)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "138c5e18",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:48.307517Z",
     "iopub.status.busy": "2022-04-16T08:51:48.306846Z",
     "iopub.status.idle": "2022-04-16T08:51:48.663811Z",
     "shell.execute_reply": "2022-04-16T08:51:48.663310Z",
     "shell.execute_reply.started": "2022-04-16T05:04:58.097693Z"
    },
    "id": "uoKZ8VeBOWzJ",
    "papermill": {
     "duration": 0.425961,
     "end_time": "2022-04-16T08:51:48.663947",
     "exception": false,
     "start_time": "2022-04-16T08:51:48.237986",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "train, test = train_test_split(dataset, test_size=0.1, random_state=44)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1a731254",
   "metadata": {
    "id": "3mReEzwzPr9x",
    "papermill": {
     "duration": 0.058481,
     "end_time": "2022-04-16T08:51:48.781859",
     "exception": false,
     "start_time": "2022-04-16T08:51:48.723378",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Results "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "75557dec",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:48.905172Z",
     "iopub.status.busy": "2022-04-16T08:51:48.904302Z",
     "iopub.status.idle": "2022-04-16T08:51:48.907708Z",
     "shell.execute_reply": "2022-04-16T08:51:48.908241Z",
     "shell.execute_reply.started": "2022-04-16T05:04:58.488558Z"
    },
    "id": "aECLgCaePjaU",
    "outputId": "7627359b-9e41-4fa5-8ffc-daaba24acbd4",
    "papermill": {
     "duration": 0.067013,
     "end_time": "2022-04-16T08:51:48.908415",
     "exception": false,
     "start_time": "2022-04-16T08:51:48.841402",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set length: 1.44M examples\n",
      "Test set length: 0.16M examples\n"
     ]
    }
   ],
   "source": [
    "print(f\"Training set length: {len(train)/1e6}M examples\")\n",
    "print(f\"Test set length: {len(test)/1e6}M examples\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "36b4f702",
   "metadata": {
    "id": "lG9lPGZgQJ7r",
    "papermill": {
     "duration": 0.059257,
     "end_time": "2022-04-16T08:51:49.027338",
     "exception": false,
     "start_time": "2022-04-16T08:51:48.968081",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Tokenization\n",
    "\n",
    "Tokenization is the process of breaking apart original text into individual pieces (tokens) for further analysis. This process is important because the meaning of the text can be interpreted through analysis of the words present in the text. \n",
    "\n",
    "[Inspired from this medium article](https://medium.com/analytics-vidhya/spacy-basics-the-importance-of-tokens-in-natural-language-processing-89a698d8a76). "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3b8503e9",
   "metadata": {
    "id": "XG5wc4ugQ18w",
    "papermill": {
     "duration": 0.070845,
     "end_time": "2022-04-16T08:51:49.157892",
     "exception": false,
     "start_time": "2022-04-16T08:51:49.087047",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Define the tokenizer\n",
    "\n",
    "This class allows to vectorize a text corpus, by turning each text into either a sequence of integers (each integer being the index of a token in a dictionary) or into a vector where the coefficient for each token could be binary, based on word count, based on tf-idf... [[source]](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/text/Tokenizer). \n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "1aafb18e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:51:49.395082Z",
     "iopub.status.busy": "2022-04-16T08:51:49.394305Z",
     "iopub.status.idle": "2022-04-16T08:52:15.265707Z",
     "shell.execute_reply": "2022-04-16T08:52:15.265187Z",
     "shell.execute_reply.started": "2022-04-16T05:04:58.49484Z"
    },
    "id": "tGx6CvE8P2Tv",
    "papermill": {
     "duration": 26.001596,
     "end_time": "2022-04-16T08:52:15.265859",
     "exception": false,
     "start_time": "2022-04-16T08:51:49.264263",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Initialize the tokenizer\n",
    "tokenizer = Tokenizer()\n",
    "# Updates the internal vocabulary based on our tweet contents\n",
    "tokenizer.fit_on_texts(train.content)\n",
    "# Setup the vocabular size based on the tokenizer results\n",
    "vocab_size = len(tokenizer.word_index)+1"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "182db9ec",
   "metadata": {
    "papermill": {
     "duration": 0.058863,
     "end_time": "2022-04-16T08:52:15.384422",
     "exception": false,
     "start_time": "2022-04-16T08:52:15.325559",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Results "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "b940a840",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:52:15.514261Z",
     "iopub.status.busy": "2022-04-16T08:52:15.513568Z",
     "iopub.status.idle": "2022-04-16T08:52:15.516429Z",
     "shell.execute_reply": "2022-04-16T08:52:15.517033Z",
     "shell.execute_reply.started": "2022-04-16T05:05:24.417521Z"
    },
    "papermill": {
     "duration": 0.07376,
     "end_time": "2022-04-16T08:52:15.517202",
     "exception": false,
     "start_time": "2022-04-16T08:52:15.443442",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Word index length: 268198\n",
      "Some words: ['i', 'to', 'the', 'a', 'my', 'it', 'and', 'you', 'is', 'in']\n"
     ]
    }
   ],
   "source": [
    "print(f\"Word index length: {len(tokenizer.word_index)}\")\n",
    "print(f\"Some words: {list(tokenizer.word_index.keys())[0:10]}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b6761c73",
   "metadata": {
    "id": "Pbm6bM-7RIuW",
    "papermill": {
     "duration": 0.061841,
     "end_time": "2022-04-16T08:52:15.639468",
     "exception": false,
     "start_time": "2022-04-16T08:52:15.577627",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Get the max token length"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "e5005766",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:52:15.788719Z",
     "iopub.status.busy": "2022-04-16T08:52:15.787725Z",
     "iopub.status.idle": "2022-04-16T08:52:17.588779Z",
     "shell.execute_reply": "2022-04-16T08:52:17.588175Z",
     "shell.execute_reply.started": "2022-04-16T05:05:24.435006Z"
    },
    "id": "Wb1Y8hnVRR-U",
    "outputId": "48614b23-8caf-403c-ce10-d49cb521ce94",
    "papermill": {
     "duration": 1.88137,
     "end_time": "2022-04-16T08:52:17.588942",
     "exception": false,
     "start_time": "2022-04-16T08:52:15.707572",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum token length: 53\n"
     ]
    }
   ],
   "source": [
    "# Intialize the mex length to the first tweet length\n",
    "max_length = len(dataset[\"content\"][0].split())\n",
    "\n",
    "# Loop through the tweets\n",
    "for tweet in dataset[\"content\"]: \n",
    "    # Get each tweet's length\n",
    "    length = len(tweet.split())\n",
    "    # Update the max length if greater\n",
    "    if length > max_length: \n",
    "        max_length = length\n",
    "\n",
    "print(f\"Maximum token length: {max_length}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e5079b0d",
   "metadata": {
    "id": "RoV-mJttTZvW",
    "papermill": {
     "duration": 0.067754,
     "end_time": "2022-04-16T08:52:17.718719",
     "exception": false,
     "start_time": "2022-04-16T08:52:17.650965",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Define sequences\n",
    "\n",
    "The Tensorlow \"texts_to_sequences\" method transforms each text in texts to a sequence of integers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "506567cf",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:52:17.873522Z",
     "iopub.status.busy": "2022-04-16T08:52:17.868245Z",
     "iopub.status.idle": "2022-04-16T08:52:42.506262Z",
     "shell.execute_reply": "2022-04-16T08:52:42.505775Z",
     "shell.execute_reply.started": "2022-04-16T05:05:26.202359Z"
    },
    "id": "BH5ehzZzRdB0",
    "papermill": {
     "duration": 24.719601,
     "end_time": "2022-04-16T08:52:42.506395",
     "exception": false,
     "start_time": "2022-04-16T08:52:17.786794",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Training set\n",
    "sequences_train = tokenizer.texts_to_sequences(train.content)\n",
    "# Test set\n",
    "sequences_test = tokenizer.texts_to_sequences(test.content)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "db548dc9",
   "metadata": {
    "id": "5qBX1xc-TpbU",
    "papermill": {
     "duration": 0.05917,
     "end_time": "2022-04-16T08:52:42.625580",
     "exception": false,
     "start_time": "2022-04-16T08:52:42.566410",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Define the training and test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "cb32fb8c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:52:42.804511Z",
     "iopub.status.busy": "2022-04-16T08:52:42.794564Z",
     "iopub.status.idle": "2022-04-16T08:52:54.971863Z",
     "shell.execute_reply": "2022-04-16T08:52:54.972723Z",
     "shell.execute_reply.started": "2022-04-16T05:05:50.842148Z"
    },
    "id": "UoiSTXzfTnLl",
    "outputId": "cab2975a-b088-4309-f884-88b39eb5a926",
    "papermill": {
     "duration": 12.288199,
     "end_time": "2022-04-16T08:52:54.972916",
     "exception": false,
     "start_time": "2022-04-16T08:52:42.684717",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training test shape : (1440000, 53)\n"
     ]
    }
   ],
   "source": [
    "# Pad the training sequences to the same length after the sequence\n",
    "X_train = pad_sequences(sequences_train, maxlen=max_length, padding=\"post\")\n",
    "# Pad the test sequences to the same length after the sequence\n",
    "X_test = pad_sequences(sequences_test, maxlen=max_length, padding=\"post\")\n",
    "\n",
    "# Setup the training target\n",
    "y_train = train.target.values\n",
    "# Setup the test target values\n",
    "y_test = test.target.values\n",
    "\n",
    "print(f\"Training test shape : {X_train.shape}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d49d86db",
   "metadata": {
    "id": "ryPd1T9BVkBa",
    "papermill": {
     "duration": 0.065176,
     "end_time": "2022-04-16T08:52:55.100143",
     "exception": false,
     "start_time": "2022-04-16T08:52:55.034967",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Embedding with Word Embeddings (GloVe)\n",
    "\n",
    "- Embedding is an approach that allows us to give a numeric representation of words. The Word Embeddings method setup words with similar meaning to have a similar representation. [Learn more here](https://www.geeksforgeeks.org/word-embeddings-in-nlp/)\n",
    "- To sum up briefly, the GloVe dataset is a model for distributed word representation pre-trained on a very large corpus. Our GloVe dataset contains 100-dimensional pretrained words, with each row containing the actual word followed by its 100-dimensional representation. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0721eeb5",
   "metadata": {
    "id": "AXKpPqNFWK3n",
    "papermill": {
     "duration": 0.107139,
     "end_time": "2022-04-16T08:52:55.306875",
     "exception": false,
     "start_time": "2022-04-16T08:52:55.199736",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Setup embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "287a3ba5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:52:55.510914Z",
     "iopub.status.busy": "2022-04-16T08:52:55.510119Z",
     "iopub.status.idle": "2022-04-16T08:52:55.516572Z",
     "shell.execute_reply": "2022-04-16T08:52:55.515930Z",
     "shell.execute_reply.started": "2022-04-16T05:08:01.532439Z"
    },
    "id": "cJqPXut3VWWh",
    "papermill": {
     "duration": 0.110356,
     "end_time": "2022-04-16T08:52:55.516740",
     "exception": false,
     "start_time": "2022-04-16T08:52:55.406384",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "embeddings_dictionary = dict()\n",
    "embedding_dimension = 100\n",
    "glove_file = open(\"/kaggle/input/twitter-sentiment-analysis-and-word-embeddings/Dataset/Dataset/word_embeddings.txt\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ba817dd0",
   "metadata": {
    "id": "TUEMLeF-Wops",
    "papermill": {
     "duration": 0.096635,
     "end_time": "2022-04-16T08:52:55.711482",
     "exception": false,
     "start_time": "2022-04-16T08:52:55.614847",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Apply word embedding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "7c0cd65b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:52:55.840280Z",
     "iopub.status.busy": "2022-04-16T08:52:55.839742Z",
     "iopub.status.idle": "2022-04-16T08:53:13.610470Z",
     "shell.execute_reply": "2022-04-16T08:53:13.609931Z",
     "shell.execute_reply.started": "2022-04-16T05:08:03.318175Z"
    },
    "id": "ISKnOvqPWHwc",
    "papermill": {
     "duration": 17.836787,
     "end_time": "2022-04-16T08:53:13.610593",
     "exception": false,
     "start_time": "2022-04-16T08:52:55.773806",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Iterate through the glove file\n",
    "for line in glove_file:\n",
    "    # Split each line\n",
    "    records = line.split()\n",
    "    # Get the actual word\n",
    "    word = records[0]\n",
    "    # Get the dimensional representation\n",
    "    representation = np.asarray(records[1:], dtype=\"float32\")\n",
    "    # Update the words dictionary \n",
    "    embeddings_dictionary[word] = representation\n",
    "    \n",
    "# Close the file\n",
    "glove_file.close()\n",
    "\n",
    "# Initialize the embedding_matrix\n",
    "embeddings_matrix = np.zeros((vocab_size, embedding_dimension))\n",
    "\n",
    "# Iterate through the tokenizer words list\n",
    "for word, index in tokenizer.word_index.items():\n",
    "    # Get the word representation\n",
    "    embedding_vector = embeddings_dictionary.get(word)\n",
    "    # Update the word representation if it exist in our dictionary\n",
    "    if embedding_vector is not None:\n",
    "        embeddings_matrix[index] = embedding_vector"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a6376726",
   "metadata": {
    "id": "f1O7nHkcX98b",
    "papermill": {
     "duration": 0.061793,
     "end_time": "2022-04-16T08:53:13.733693",
     "exception": false,
     "start_time": "2022-04-16T08:53:13.671900",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Define the model 🔧"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0d415ee9",
   "metadata": {
    "id": "dzmRd54PYA1R",
    "papermill": {
     "duration": 0.064235,
     "end_time": "2022-04-16T08:53:13.863839",
     "exception": false,
     "start_time": "2022-04-16T08:53:13.799604",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Set the training parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "8b7788ae",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:53:13.993898Z",
     "iopub.status.busy": "2022-04-16T08:53:13.992957Z",
     "iopub.status.idle": "2022-04-16T08:53:13.995489Z",
     "shell.execute_reply": "2022-04-16T08:53:13.994973Z",
     "shell.execute_reply.started": "2022-04-16T05:08:20.398763Z"
    },
    "id": "DPEKs1DiWRQq",
    "papermill": {
     "duration": 0.069722,
     "end_time": "2022-04-16T08:53:13.995597",
     "exception": false,
     "start_time": "2022-04-16T08:53:13.925875",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "num_epochs = 50\n",
    "batch_size = 1000"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d07c8205",
   "metadata": {
    "id": "u-MBH9dkYgGH",
    "papermill": {
     "duration": 0.060106,
     "end_time": "2022-04-16T08:53:14.116328",
     "exception": false,
     "start_time": "2022-04-16T08:53:14.056222",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Define the embedding layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "587b1cba",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:53:14.251463Z",
     "iopub.status.busy": "2022-04-16T08:53:14.250768Z",
     "iopub.status.idle": "2022-04-16T08:53:14.263638Z",
     "shell.execute_reply": "2022-04-16T08:53:14.263244Z",
     "shell.execute_reply.started": "2022-04-16T05:08:20.403973Z"
    },
    "id": "v55TleKdYEvX",
    "papermill": {
     "duration": 0.07873,
     "end_time": "2022-04-16T08:53:14.263774",
     "exception": false,
     "start_time": "2022-04-16T08:53:14.185044",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "embedding_layer = Embedding(\n",
    "                            vocab_size, \n",
    "                            embedding_dimension, \n",
    "                            input_length=max_length, \n",
    "                            weights=[embeddings_matrix], \n",
    "                            trainable=False\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3961311d",
   "metadata": {
    "id": "nFD3_-1WeZ0k",
    "papermill": {
     "duration": 0.070517,
     "end_time": "2022-04-16T08:53:14.396342",
     "exception": false,
     "start_time": "2022-04-16T08:53:14.325825",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Define the callback : Early Stopping\n",
    "\n",
    "We setup the Early Stopping callback to stop the training process if the **validation accuracy** stops increasing after 10 epochs, and restore the weights which minimize the validation accuracy. \n",
    "\n",
    "Important parameters :\n",
    "\n",
    "- **Monitor :** keras will stop training when the monitored metric has stopped improving. The direction (minimze or maximize) is automatically infered from the monitored quantity name, although it remains adjustable with the \"mode\" parameter\n",
    "- **Patience :** the number of epochs without improvement after which the callback will be triggered (i.e, the model will stop training)\n",
    "- **Restore best weights :** tells the model to restore the weights from the \"best\" epoch according to the monitored value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "edb90a15",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:53:14.540545Z",
     "iopub.status.busy": "2022-04-16T08:53:14.538794Z",
     "iopub.status.idle": "2022-04-16T08:53:14.541234Z",
     "shell.execute_reply": "2022-04-16T08:53:14.541698Z",
     "shell.execute_reply.started": "2022-04-16T05:08:20.426792Z"
    },
    "id": "o1aTkm9geZKZ",
    "papermill": {
     "duration": 0.076227,
     "end_time": "2022-04-16T08:53:14.541840",
     "exception": false,
     "start_time": "2022-04-16T08:53:14.465613",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "stop_early = tf.keras.callbacks.EarlyStopping(\n",
    "    monitor=\"val_accuracy\", \n",
    "    patience=10, \n",
    "    mode =\"max\", \n",
    "    verbose=2, \n",
    "    restore_best_weights=True\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "44260216",
   "metadata": {
    "id": "_xh5SCzMYnI6",
    "papermill": {
     "duration": 0.06597,
     "end_time": "2022-04-16T08:53:14.675777",
     "exception": false,
     "start_time": "2022-04-16T08:53:14.609807",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Create the model\n",
    "\n",
    "Architecture : \n",
    "\n",
    "### The embedding layer defined before\n",
    "\n",
    "### 3 Bidirectional LSTM layers followed by dropout layers\n",
    "\n",
    "To sum up : \n",
    "- **LSTM layers** are an RNN architecture used to keep track of arbitrary long-term dependencies in the input sequences. The bidirectional version fits well in our context where each sequence (word) can be related to the previous or the next one to grasp the full meaning of the sentence.\n",
    "- **Dropout** prevents overfitting by randomly dropping neurons\n",
    "\n",
    "### Two fully-connected (dense) layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "fdddbd5c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:53:15.006155Z",
     "iopub.status.busy": "2022-04-16T08:53:15.005314Z",
     "iopub.status.idle": "2022-04-16T08:53:19.194005Z",
     "shell.execute_reply": "2022-04-16T08:53:19.194460Z",
     "shell.execute_reply.started": "2022-04-16T05:08:20.433484Z"
    },
    "id": "zlQFKnznacrY",
    "papermill": {
     "duration": 4.453287,
     "end_time": "2022-04-16T08:53:19.194615",
     "exception": false,
     "start_time": "2022-04-16T08:53:14.741328",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-04-16 08:53:14.816232: I tensorflow/compiler/jit/xla_cpu_device.cc:41] Not creating XLA devices, tf_xla_enable_xla_devices not set\n",
      "2022-04-16 08:53:14.819440: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcuda.so.1\n",
      "2022-04-16 08:53:14.882999: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:14.883738: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
      "pciBusID: 0000:00:04.0 name: Tesla P100-PCIE-16GB computeCapability: 6.0\n",
      "coreClock: 1.3285GHz coreCount: 56 deviceMemorySize: 15.90GiB deviceMemoryBandwidth: 681.88GiB/s\n",
      "2022-04-16 08:53:14.883800: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0\n",
      "2022-04-16 08:53:14.912651: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.11\n",
      "2022-04-16 08:53:14.912764: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.11\n",
      "2022-04-16 08:53:14.928705: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
      "2022-04-16 08:53:14.953579: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
      "2022-04-16 08:53:14.989601: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
      "2022-04-16 08:53:14.997832: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.11\n",
      "2022-04-16 08:53:15.000742: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.8\n",
      "2022-04-16 08:53:15.000926: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:15.001719: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:15.003437: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
      "2022-04-16 08:53:15.004877: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 AVX512F FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2022-04-16 08:53:15.005147: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set\n",
      "2022-04-16 08:53:15.005326: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:15.006230: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
      "pciBusID: 0000:00:04.0 name: Tesla P100-PCIE-16GB computeCapability: 6.0\n",
      "coreClock: 1.3285GHz coreCount: 56 deviceMemorySize: 15.90GiB deviceMemoryBandwidth: 681.88GiB/s\n",
      "2022-04-16 08:53:15.006309: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0\n",
      "2022-04-16 08:53:15.006363: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.11\n",
      "2022-04-16 08:53:15.006392: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.11\n",
      "2022-04-16 08:53:15.006418: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
      "2022-04-16 08:53:15.006452: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
      "2022-04-16 08:53:15.006478: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
      "2022-04-16 08:53:15.006505: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.11\n",
      "2022-04-16 08:53:15.006531: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.8\n",
      "2022-04-16 08:53:15.006623: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:15.007298: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:15.007860: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
      "2022-04-16 08:53:15.009246: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0\n",
      "2022-04-16 08:53:16.420374: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1261] Device interconnect StreamExecutor with strength 1 edge matrix:\n",
      "2022-04-16 08:53:16.420420: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1267]      0 \n",
      "2022-04-16 08:53:16.420431: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1280] 0:   N \n",
      "2022-04-16 08:53:16.422463: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:16.423286: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:16.423971: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2022-04-16 08:53:16.424551: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1406] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14957 MB memory) -> physical GPU (device: 0, name: Tesla P100-PCIE-16GB, pci bus id: 0000:00:04.0, compute capability: 6.0)\n"
     ]
    }
   ],
   "source": [
    "model = Sequential([\n",
    "        embedding_layer,\n",
    "        tf.keras.layers.Bidirectional(LSTM(units=128, return_sequences=True)),\n",
    "        tf.keras.layers.Dropout(0.2),\n",
    "        tf.keras.layers.Bidirectional(LSTM(units=128, return_sequences=True)),\n",
    "        tf.keras.layers.Dropout(0.2),\n",
    "        tf.keras.layers.Bidirectional(LSTM(units=128)),\n",
    "        tf.keras.layers.Dropout(0.2),\n",
    "        tf.keras.layers.Dense(64, activation=\"relu\"),\n",
    "        tf.keras.layers.Dense(1, activation=\"sigmoid\"),\n",
    "    ])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "90188cbc",
   "metadata": {
    "id": "GBLDT_nyZuD3",
    "papermill": {
     "duration": 0.061095,
     "end_time": "2022-04-16T08:53:19.317718",
     "exception": false,
     "start_time": "2022-04-16T08:53:19.256623",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Compile the model\n",
    "\n",
    "> Notes : The binary cross entropy loss suits well to our binary classification case (two targets : positive or negative)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "4d15add4",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:53:19.449467Z",
     "iopub.status.busy": "2022-04-16T08:53:19.448952Z",
     "iopub.status.idle": "2022-04-16T08:53:19.455756Z",
     "shell.execute_reply": "2022-04-16T08:53:19.455292Z",
     "shell.execute_reply.started": "2022-04-16T05:09:39.114642Z"
    },
    "id": "cs-ego69YrAj",
    "papermill": {
     "duration": 0.077637,
     "end_time": "2022-04-16T08:53:19.455869",
     "exception": false,
     "start_time": "2022-04-16T08:53:19.378232",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model.compile(\n",
    "       optimizer = Adam(learning_rate=0.001),\n",
    "       loss=\"binary_crossentropy\", \n",
    "       metrics=[\"accuracy\"]\n",
    "  )"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "df849ba5",
   "metadata": {
    "id": "3QFsvRwkbent",
    "papermill": {
     "duration": 0.060127,
     "end_time": "2022-04-16T08:53:19.576235",
     "exception": false,
     "start_time": "2022-04-16T08:53:19.516108",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Model summary "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "11d0f61e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:53:19.706397Z",
     "iopub.status.busy": "2022-04-16T08:53:19.700643Z",
     "iopub.status.idle": "2022-04-16T08:53:19.710396Z",
     "shell.execute_reply": "2022-04-16T08:53:19.709990Z",
     "shell.execute_reply.started": "2022-04-16T05:09:41.245225Z"
    },
    "id": "CQXLnEwgZzy6",
    "outputId": "98606ea4-b661-474d-b409-e4ae837b4fd4",
    "papermill": {
     "duration": 0.073901,
     "end_time": "2022-04-16T08:53:19.710513",
     "exception": false,
     "start_time": "2022-04-16T08:53:19.636612",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, 53, 100)           26819900  \n",
      "_________________________________________________________________\n",
      "bidirectional (Bidirectional (None, 53, 256)           234496    \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 53, 256)           0         \n",
      "_________________________________________________________________\n",
      "bidirectional_1 (Bidirection (None, 53, 256)           394240    \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 53, 256)           0         \n",
      "_________________________________________________________________\n",
      "bidirectional_2 (Bidirection (None, 256)               394240    \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 64)                16448     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 65        \n",
      "=================================================================\n",
      "Total params: 27,859,389\n",
      "Trainable params: 1,039,489\n",
      "Non-trainable params: 26,819,900\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ea2ef6bf",
   "metadata": {
    "papermill": {
     "duration": 0.062895,
     "end_time": "2022-04-16T08:53:19.834291",
     "exception": false,
     "start_time": "2022-04-16T08:53:19.771396",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Plot the model architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "df9feae7",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:53:19.964150Z",
     "iopub.status.busy": "2022-04-16T08:53:19.963270Z",
     "iopub.status.idle": "2022-04-16T08:53:20.853047Z",
     "shell.execute_reply": "2022-04-16T08:53:20.852513Z",
     "shell.execute_reply.started": "2022-04-16T05:09:53.141253Z"
    },
    "id": "nsbS_JxcfnkB",
    "outputId": "2dc9dfda-f615-41bb-c01a-60aa665eb9bb",
    "papermill": {
     "duration": 0.95701,
     "end_time": "2022-04-16T08:53:20.853166",
     "exception": false,
     "start_time": "2022-04-16T08:53:19.896156",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.keras.utils.plot_model(model, show_shapes=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bacd85c1",
   "metadata": {
    "id": "U1wMN2Lcbq3V",
    "papermill": {
     "duration": 0.064467,
     "end_time": "2022-04-16T08:53:20.982771",
     "exception": false,
     "start_time": "2022-04-16T08:53:20.918304",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Train the model ⚗️"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "3e6739c8",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T08:53:21.116894Z",
     "iopub.status.busy": "2022-04-16T08:53:21.116271Z",
     "iopub.status.idle": "2022-04-16T09:41:50.637721Z",
     "shell.execute_reply": "2022-04-16T09:41:50.638597Z"
    },
    "id": "ExFPK4f3bhay",
    "outputId": "cdc7f500-6ba1-45a7-d1cd-3ae4a8540444",
    "papermill": {
     "duration": 2909.591987,
     "end_time": "2022-04-16T09:41:50.638814",
     "exception": false,
     "start_time": "2022-04-16T08:53:21.046827",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-04-16 08:53:21.385426: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:116] None of the MLIR optimization passes are enabled (registered 2)\n",
      "2022-04-16 08:53:21.394585: I tensorflow/core/platform/profile_utils/cpu_utils.cc:112] CPU Frequency: 2000129999 Hz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-04-16 08:53:26.725844: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.11\n",
      "2022-04-16 08:53:27.468088: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.11\n",
      "2022-04-16 08:53:27.530210: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.8\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1440/1440 [==============================] - 175s 113ms/step - loss: 0.5146 - accuracy: 0.7414 - val_loss: 0.4350 - val_accuracy: 0.7969\n",
      "Epoch 2/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.4265 - accuracy: 0.8024 - val_loss: 0.4169 - val_accuracy: 0.8074\n",
      "Epoch 3/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.4044 - accuracy: 0.8146 - val_loss: 0.4000 - val_accuracy: 0.8159\n",
      "Epoch 4/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3889 - accuracy: 0.8231 - val_loss: 0.3899 - val_accuracy: 0.8228\n",
      "Epoch 5/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3773 - accuracy: 0.8297 - val_loss: 0.3883 - val_accuracy: 0.8249\n",
      "Epoch 6/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3663 - accuracy: 0.8360 - val_loss: 0.3798 - val_accuracy: 0.8286\n",
      "Epoch 7/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3580 - accuracy: 0.8396 - val_loss: 0.3834 - val_accuracy: 0.8291\n",
      "Epoch 8/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3494 - accuracy: 0.8447 - val_loss: 0.3821 - val_accuracy: 0.8309\n",
      "Epoch 9/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3410 - accuracy: 0.8488 - val_loss: 0.3829 - val_accuracy: 0.8287\n",
      "Epoch 10/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3321 - accuracy: 0.8537 - val_loss: 0.3840 - val_accuracy: 0.8303\n",
      "Epoch 11/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3236 - accuracy: 0.8574 - val_loss: 0.3821 - val_accuracy: 0.8293\n",
      "Epoch 12/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3158 - accuracy: 0.8619 - val_loss: 0.3840 - val_accuracy: 0.8290\n",
      "Epoch 13/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3082 - accuracy: 0.8653 - val_loss: 0.3921 - val_accuracy: 0.8291\n",
      "Epoch 14/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.3012 - accuracy: 0.8690 - val_loss: 0.3993 - val_accuracy: 0.8261\n",
      "Epoch 15/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.2929 - accuracy: 0.8730 - val_loss: 0.4076 - val_accuracy: 0.8263\n",
      "Epoch 16/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.2862 - accuracy: 0.8761 - val_loss: 0.4087 - val_accuracy: 0.8267\n",
      "Epoch 17/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.2807 - accuracy: 0.8794 - val_loss: 0.4195 - val_accuracy: 0.8240\n",
      "Epoch 18/50\n",
      "1440/1440 [==============================] - 161s 112ms/step - loss: 0.2734 - accuracy: 0.8826 - val_loss: 0.4155 - val_accuracy: 0.8233\n",
      "Restoring model weights from the end of the best epoch.\n",
      "Epoch 00018: early stopping\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(\n",
    "    X_train,\n",
    "    y_train,\n",
    "    batch_size=batch_size,\n",
    "    epochs=num_epochs,\n",
    "    verbose=1, \n",
    "    validation_data=(X_test,y_test),\n",
    "    callbacks = [stop_early]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "64d47dc9",
   "metadata": {
    "papermill": {
     "duration": 6.923058,
     "end_time": "2022-04-16T09:42:05.054670",
     "exception": false,
     "start_time": "2022-04-16T09:41:58.131612",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Save the model  💾"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7041be3f",
   "metadata": {
    "papermill": {
     "duration": 7.096723,
     "end_time": "2022-04-16T09:42:18.891309",
     "exception": false,
     "start_time": "2022-04-16T09:42:11.794586",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    " ## Define the text vectorization layer for inferences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "e7bd93e5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:42:33.365206Z",
     "iopub.status.busy": "2022-04-16T09:42:33.364338Z",
     "iopub.status.idle": "2022-04-16T09:42:33.366505Z",
     "shell.execute_reply": "2022-04-16T09:42:33.366905Z"
    },
    "papermill": {
     "duration": 7.435223,
     "end_time": "2022-04-16T09:42:33.367058",
     "exception": false,
     "start_time": "2022-04-16T09:42:25.931835",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Setup the vocabulary\n",
    "vocabulary = list(tokenizer.word_index.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "29414b4e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:42:47.268881Z",
     "iopub.status.busy": "2022-04-16T09:42:47.267192Z",
     "iopub.status.idle": "2022-04-16T09:42:47.271343Z",
     "shell.execute_reply": "2022-04-16T09:42:47.270683Z"
    },
    "papermill": {
     "duration": 6.810867,
     "end_time": "2022-04-16T09:42:47.271500",
     "exception": false,
     "start_time": "2022-04-16T09:42:40.460633",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model input shape: (None, 53)\n"
     ]
    }
   ],
   "source": [
    "# Get the input shape\n",
    "input_shape = tf.keras.Input(shape=(max_length))\n",
    "shape = input_shape.shape\n",
    "print(f\"Model input shape: {shape}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "d198e28b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:43:01.777164Z",
     "iopub.status.busy": "2022-04-16T09:43:01.776348Z",
     "iopub.status.idle": "2022-04-16T09:43:19.164832Z",
     "shell.execute_reply": "2022-04-16T09:43:19.165781Z"
    },
    "papermill": {
     "duration": 25.000314,
     "end_time": "2022-04-16T09:43:19.166055",
     "exception": false,
     "start_time": "2022-04-16T09:42:54.165741",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Create the text vectorization layer\n",
    "vectorize_layer = TextVectorization (\n",
    "    # Standardization to apply to the input text\n",
    "    standardize=\"lower_and_strip_punctuation\",\n",
    "    # Maximum size of the vocabulary for this layer\n",
    "    max_tokens=vocab_size+1,\n",
    "    # Output of the layer \n",
    "    output_mode=\"int\",\n",
    "    output_sequence_length=max_length, \n",
    "    vocabulary=vocabulary\n",
    ")\n",
    "\n",
    "# Adapt to our training content\n",
    "vectorize_layer.adapt(train[\"content\"].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e78b5ebf",
   "metadata": {
    "papermill": {
     "duration": 7.595169,
     "end_time": "2022-04-16T09:43:33.763238",
     "exception": false,
     "start_time": "2022-04-16T09:43:26.168069",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Define the export model\n",
    "\n",
    "Export model : **Input Layer ---> Vectorization Layer ---> Our trained model ---> The activation layer (score)**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "f8d5a69e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:43:47.559310Z",
     "iopub.status.busy": "2022-04-16T09:43:47.558479Z",
     "iopub.status.idle": "2022-04-16T09:43:48.711451Z",
     "shell.execute_reply": "2022-04-16T09:43:48.710996Z"
    },
    "papermill": {
     "duration": 8.173661,
     "end_time": "2022-04-16T09:43:48.711584",
     "exception": false,
     "start_time": "2022-04-16T09:43:40.537923",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "export_model = tf.keras.Sequential([\n",
    "    tf.keras.layers.InputLayer(input_shape=(1,), dtype=tf.string),\n",
    "    vectorize_layer,\n",
    "    model,\n",
    "    tf.keras.layers.Activation(\"sigmoid\")\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f7dc4ecb",
   "metadata": {
    "papermill": {
     "duration": 6.87182,
     "end_time": "2022-04-16T09:44:02.756131",
     "exception": false,
     "start_time": "2022-04-16T09:43:55.884311",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Compile the export model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "3307f283",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:44:17.477827Z",
     "iopub.status.busy": "2022-04-16T09:44:17.477204Z",
     "iopub.status.idle": "2022-04-16T09:44:17.483070Z",
     "shell.execute_reply": "2022-04-16T09:44:17.482565Z"
    },
    "papermill": {
     "duration": 7.14374,
     "end_time": "2022-04-16T09:44:17.483187",
     "exception": false,
     "start_time": "2022-04-16T09:44:10.339447",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "export_model.compile(\n",
    "       optimizer = Adam(learning_rate = 0.001),\n",
    "       loss=\"binary_crossentropy\", \n",
    "       metrics=[\"accuracy\"]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bb19fea3",
   "metadata": {
    "papermill": {
     "duration": 6.966996,
     "end_time": "2022-04-16T09:44:31.219200",
     "exception": false,
     "start_time": "2022-04-16T09:44:24.252204",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Save the export model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "638d8848",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:44:45.479879Z",
     "iopub.status.busy": "2022-04-16T09:44:45.478649Z",
     "iopub.status.idle": "2022-04-16T09:45:26.397018Z",
     "shell.execute_reply": "2022-04-16T09:45:26.396302Z"
    },
    "papermill": {
     "duration": 47.641426,
     "end_time": "2022-04-16T09:45:26.397207",
     "exception": false,
     "start_time": "2022-04-16T09:44:38.755781",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-04-16 09:45:01.131832: W tensorflow/python/util/util.cc:348] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.\n"
     ]
    }
   ],
   "source": [
    "export_model.save(\"TSA_model_v4\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "86d5621f",
   "metadata": {
    "papermill": {
     "duration": 6.751322,
     "end_time": "2022-04-16T09:45:40.225279",
     "exception": false,
     "start_time": "2022-04-16T09:45:33.473957",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Plot the results 📊"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "d6d02d58",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:45:54.827125Z",
     "iopub.status.busy": "2022-04-16T09:45:54.826401Z",
     "iopub.status.idle": "2022-04-16T09:45:54.829716Z",
     "shell.execute_reply": "2022-04-16T09:45:54.830207Z"
    },
    "papermill": {
     "duration": 7.034674,
     "end_time": "2022-04-16T09:45:54.830353",
     "exception": false,
     "start_time": "2022-04-16T09:45:47.795679",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of epochs: 18\n"
     ]
    }
   ],
   "source": [
    "number_of_epochs = len(history.history[\"loss\"])\n",
    "print(f\"Number of epochs: {number_of_epochs}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "beb9a50b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:46:08.803420Z",
     "iopub.status.busy": "2022-04-16T09:46:08.802546Z",
     "iopub.status.idle": "2022-04-16T09:46:09.137863Z",
     "shell.execute_reply": "2022-04-16T09:46:09.138334Z"
    },
    "papermill": {
     "duration": 7.39329,
     "end_time": "2022-04-16T09:46:09.138493",
     "exception": false,
     "start_time": "2022-04-16T09:46:01.745203",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.8309249877929688\n",
      "Minimum loss: 0.3820512592792511\n"
     ]
    }
   ],
   "source": [
    "# Get the accuracy results for each epoch\n",
    "acc = history.history[\"accuracy\"]\n",
    "val_acc = history.history[\"val_accuracy\"]\n",
    "# Get the loss results for each epoch\n",
    "loss = history.history[\"loss\"]\n",
    "val_loss = history.history[\"val_loss\"]\n",
    "# Get the epochs count\n",
    "epochs_range = range(len(history.history[\"loss\"]))\n",
    "\n",
    "# Create the plot area\n",
    "plt.figure(figsize=(14, 8))\n",
    "\n",
    "# Setup the first colum : Accuracy plot\n",
    "plt.subplot(1, 2, 1)\n",
    "# Plot the training and validation accuracy for these epochs\n",
    "plt.plot(epochs_range, acc, label=\"Training Accuracy\")\n",
    "plt.plot(epochs_range, val_acc, label=\"Validation Accuracy\")\n",
    "# Setup the legend\n",
    "plt.legend(loc=\"lower right\")\n",
    "plt.title(\"Training and Validation Accuracy\")\n",
    "\n",
    "# Setup the second colum : Loss plot\n",
    "plt.subplot(1, 2, 2)\n",
    "# Plot the training and validation loss for these epochs\n",
    "plt.plot(epochs_range, loss, label=\"Training Loss\")\n",
    "plt.plot(epochs_range, val_loss, label=\"Validation Loss\")\n",
    "# Setup the legend\n",
    "plt.legend(loc=\"upper right\")\n",
    "plt.title(\"Training and Validation Loss\")\n",
    "\n",
    "# Show the overall plot\n",
    "plt.show()\n",
    "\n",
    "position = np.argmax(history.history[\"val_accuracy\"])\n",
    "print(f\"Maximum validation accuracy:  {history.history['val_accuracy'][position]}\\nMinimum loss: {history.history['val_loss'][position]}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "35802792",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:46:23.396322Z",
     "iopub.status.busy": "2022-04-16T09:46:23.394717Z",
     "iopub.status.idle": "2022-04-16T09:46:23.398395Z",
     "shell.execute_reply": "2022-04-16T09:46:23.397942Z"
    },
    "papermill": {
     "duration": 6.86174,
     "end_time": "2022-04-16T09:46:23.398506",
     "exception": false,
     "start_time": "2022-04-16T09:46:16.536766",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best validation accuracy : 0.8309249877929688\n"
     ]
    }
   ],
   "source": [
    "best_accuracy = max(history.history[\"val_accuracy\"])\n",
    "print(f\"Best validation accuracy : {best_accuracy}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bdf04146",
   "metadata": {
    "papermill": {
     "duration": 6.823497,
     "end_time": "2022-04-16T09:46:37.171477",
     "exception": false,
     "start_time": "2022-04-16T09:46:30.347980",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Print the classification report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "5d7507bf",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:46:51.904279Z",
     "iopub.status.busy": "2022-04-16T09:46:51.903118Z",
     "iopub.status.idle": "2022-04-16T09:47:29.113330Z",
     "shell.execute_reply": "2022-04-16T09:47:29.112872Z"
    },
    "papermill": {
     "duration": 44.890981,
     "end_time": "2022-04-16T09:47:29.113466",
     "exception": false,
     "start_time": "2022-04-16T09:46:44.222485",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.84      0.82      0.83     79800\n",
      "           1       0.82      0.84      0.83     80200\n",
      "\n",
      "    accuracy                           0.83    160000\n",
      "   macro avg       0.83      0.83      0.83    160000\n",
      "weighted avg       0.83      0.83      0.83    160000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model.predict(X_test)\n",
    "predictions = np.where(predictions<0.5, 0, 1)\n",
    "\n",
    "print(classification_report(y_test, predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "20feef8c",
   "metadata": {
    "papermill": {
     "duration": 6.868261,
     "end_time": "2022-04-16T09:47:42.983408",
     "exception": false,
     "start_time": "2022-04-16T09:47:36.115147",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "f9265ebf",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:47:57.969074Z",
     "iopub.status.busy": "2022-04-16T09:47:57.968248Z",
     "iopub.status.idle": "2022-04-16T09:47:59.584991Z",
     "shell.execute_reply": "2022-04-16T09:47:59.584514Z"
    },
    "papermill": {
     "duration": 9.476401,
     "end_time": "2022-04-16T09:47:59.585121",
     "exception": false,
     "start_time": "2022-04-16T09:47:50.108720",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Examples:  ['being sick can be really cheap when it hurts too much to eat real food plus your friends make you soup']  Type:  <class 'list'>\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[0.68351]], dtype=float32)"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sample sentence\n",
    "examples = [\"being sick can be really cheap when it hurts too much to eat real food plus your friends make you soup\"]\n",
    "# Display the example\n",
    "print(\"Examples: \", examples, \" Type: \", type(examples))\n",
    "# Predict and display the score\n",
    "export_model.predict(examples)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fe8ca422",
   "metadata": {
    "papermill": {
     "duration": 6.720327,
     "end_time": "2022-04-16T09:48:13.354073",
     "exception": false,
     "start_time": "2022-04-16T09:48:06.633746",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Tune the threshold (experimental) ⚗️"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "27d5ec25",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:48:27.491645Z",
     "iopub.status.busy": "2022-04-16T09:48:27.490768Z",
     "iopub.status.idle": "2022-04-16T09:49:03.132492Z",
     "shell.execute_reply": "2022-04-16T09:49:03.131573Z"
    },
    "papermill": {
     "duration": 42.926494,
     "end_time": "2022-04-16T09:49:03.132639",
     "exception": false,
     "start_time": "2022-04-16T09:48:20.206145",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "predictions = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2d70ef13",
   "metadata": {
    "papermill": {
     "duration": 6.996277,
     "end_time": "2022-04-16T09:49:16.933247",
     "exception": false,
     "start_time": "2022-04-16T09:49:09.936970",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Calculate the best f1 score for each threshold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "7556d7b1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:49:31.276594Z",
     "iopub.status.busy": "2022-04-16T09:49:31.275589Z",
     "iopub.status.idle": "2022-04-16T09:49:31.278456Z",
     "shell.execute_reply": "2022-04-16T09:49:31.277952Z"
    },
    "papermill": {
     "duration": 7.360362,
     "end_time": "2022-04-16T09:49:31.278576",
     "exception": false,
     "start_time": "2022-04-16T09:49:23.918214",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Generate an array from 0 to 1 with a 0.001 step\n",
    "thresholds = np.arange(0, 1, 0.001)\n",
    "thres_pos = 0.1\n",
    "f1_score = 0.001"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aff84b6a",
   "metadata": {
    "papermill": {
     "duration": 6.736458,
     "end_time": "2022-04-16T09:49:45.115532",
     "exception": false,
     "start_time": "2022-04-16T09:49:38.379074",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Positive threshold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "a76484b6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:49:59.136416Z",
     "iopub.status.busy": "2022-04-16T09:49:59.135590Z",
     "iopub.status.idle": "2022-04-16T09:53:56.616170Z",
     "shell.execute_reply": "2022-04-16T09:53:56.616619Z"
    },
    "papermill": {
     "duration": 244.525936,
     "end_time": "2022-04-16T09:53:56.616797",
     "exception": false,
     "start_time": "2022-04-16T09:49:52.090861",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/sklearn/metrics/_classification.py:1221: UndefinedMetricWarning: Precision and F-score are 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, msg_start, len(result))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best threshold : 0.423 \n",
      "Best f1_score: 0.8349145204858647\n"
     ]
    }
   ],
   "source": [
    "# Loop trough the tresholds array\n",
    "for threshold in thresholds :\n",
    "    # Get the positive predictions\n",
    "    prediction = np.where(predictions>threshold, 1, 0)\n",
    "    # Generate the classification report \n",
    "    report = classification_report(y_test, prediction, output_dict=True)\n",
    "    # Get the F1-score from the report\n",
    "    f1_score_new = report[\"1\"][\"f1-score\"]\n",
    "    # Update the final F1-score if greater\n",
    "    if f1_score_new > f1_score: \n",
    "        f1_score = f1_score_new\n",
    "        thres_pos = threshold\n",
    "\n",
    "# Display the results\n",
    "print(f\"Best threshold : {thres_pos} \\nBest f1_score: {f1_score}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3622cc85",
   "metadata": {
    "papermill": {
     "duration": 6.976372,
     "end_time": "2022-04-16T09:54:10.634588",
     "exception": false,
     "start_time": "2022-04-16T09:54:03.658216",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Negative threshold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "1105e8be",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:54:25.026192Z",
     "iopub.status.busy": "2022-04-16T09:54:25.025273Z",
     "iopub.status.idle": "2022-04-16T09:58:21.332594Z",
     "shell.execute_reply": "2022-04-16T09:58:21.333176Z"
    },
    "papermill": {
     "duration": 243.85551,
     "end_time": "2022-04-16T09:58:21.333342",
     "exception": false,
     "start_time": "2022-04-16T09:54:17.477832",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best threshold : 0.625 \n",
      "Best f1_score: 0.8340107299626078\n"
     ]
    }
   ],
   "source": [
    "thres_neg = 0.1\n",
    "f1_score_neg = 0.001\n",
    "\n",
    "# Loop trough the tresholds array\n",
    "for threshold in thresholds :\n",
    "    # Get the positive predictions\n",
    "    prediction = np.where(predictions>threshold, 1, 0)\n",
    "    # Generate the classification report \n",
    "    report = classification_report(y_test, prediction, output_dict=True)\n",
    "    # Get the F1-score from the report\n",
    "    f1_score_new = report[\"0\"][\"f1-score\"]\n",
    "    # Update the final F1-score if greater\n",
    "    if f1_score_new > f1_score_neg: \n",
    "        f1_score_neg = f1_score_new\n",
    "        thres_neg = threshold\n",
    "\n",
    "# Display the results\n",
    "print(f\"Best threshold : {thres_neg} \\nBest f1_score: {f1_score_neg}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "86c3148f",
   "metadata": {
    "papermill": {
     "duration": 7.052427,
     "end_time": "2022-04-16T09:58:35.444858",
     "exception": false,
     "start_time": "2022-04-16T09:58:28.392431",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "b3ac4a5d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-04-16T09:58:49.720781Z",
     "iopub.status.busy": "2022-04-16T09:58:49.719946Z",
     "iopub.status.idle": "2022-04-16T09:58:49.776969Z",
     "shell.execute_reply": "2022-04-16T09:58:49.777984Z",
     "shell.execute_reply.started": "2022-04-16T07:32:55.182461Z"
    },
    "papermill": {
     "duration": 7.495251,
     "end_time": "2022-04-16T09:58:49.778175",
     "exception": false,
     "start_time": "2022-04-16T09:58:42.282924",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Examples: [\"this is a very good day, don't you think so ?\"] Type: <class 'list'>\n",
      "Positive (^_^)\n"
     ]
    }
   ],
   "source": [
    "examples = [ \"this is a very good day, don't you think so ?\" ]\n",
    "\n",
    "print(f\"Examples: {examples} Type: {type(examples)}\")\n",
    "\n",
    "results = np.max(export_model.predict(examples))\n",
    "\n",
    "final_threshold = thres_neg if thres_neg>thres_pos else thres_pos\n",
    "\n",
    "if results >= final_threshold: \n",
    "    print(\"Positive (^_^)\")\n",
    "else: \n",
    "    print(\"Negative (>_<)\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.10"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 4073.093991,
   "end_time": "2022-04-16T09:58:59.509109",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2022-04-16T08:51:06.415118",
   "version": "2.3.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}