Coursework_2_Neural_Text_Classification.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Avk6evwih05Z",
        "outputId": "d9a5be29-1e3c-4418-e99c-a9ee7c7570b5"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "[nltk_data] Downloading package words to /root/nltk_data...\n",
            "[nltk_data]   Unzipping corpora/words.zip.\n",
            "[nltk_data] Downloading package punkt to /root/nltk_data...\n",
            "[nltk_data]   Unzipping tokenizers/punkt.zip.\n",
            "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]   Unzipping corpora/stopwords.zip.\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {},
          "execution_count": 2
        }
      ],
      "source": [
        "import nltk\n",
        "nltk.download('words')\n",
        "from nltk.corpus import words\n",
        "nltk.download('punkt')\n",
        "nltk.download('stopwords')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "DbcZapAjh4VU"
      },
      "outputs": [],
      "source": [
        "from nltk.tokenize import MWETokenizer, word_tokenize\n",
        "from nltk.corpus import stopwords\n",
        "from nltk.stem import SnowballStemmer\n",
        "from collections import Counter"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "4WrAcj6ih5UW"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import tqdm\n",
        "import pandas as pd\n",
        "import pathlib\n",
        "import os\n",
        "import re\n",
        "import string\n",
        "import time\n",
        "from random import random\n",
        "from collections import defaultdict\n",
        "from typing import List, Union\n",
        "import pprint\n",
        "import json\n",
        "import tensorflow as tf\n",
        "from gensim.models import Word2Vec\n",
        "from time import time"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import seaborn as sns\n",
        "import matplotlib.pyplot as plt"
      ],
      "metadata": {
        "id": "IE9URc498WH7"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gZOu6zQ4h9pF"
      },
      "source": [
        "#### Mount to Google Drive"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XEIPSlyyh6R6",
        "outputId": "d9b4dc80-d50b-4764-d89b-4c28cdb2e411"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "product_reviews_dir = 'product_reviews'\n",
        "drive_path = \"/content/drive\" + \"/My Drive/Colab Notebooks/\" + product_reviews_dir\n",
        "files = os.listdir(drive_path)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 71,
      "metadata": {
        "id": "gHO-bnjMgVbL"
      },
      "outputs": [],
      "source": [
        "def read_data(path: str) -> List[str]:\n",
        "    \"\"\"\n",
        "        Read files from a directory and then append the data of each file into a list.\n",
        "    :param path:\n",
        "    :return:\n",
        "    \"\"\"\n",
        "    files = os.listdir(path)  # to be iterated over in folder\n",
        "    content = \"\"  # to store the contents of each file in folder\n",
        "    files.remove('README.txt')\n",
        "    for filename in files:\n",
        "        file_path = f'{path}/{filename}'\n",
        "        file_extension = pathlib.Path(file_path).suffix\n",
        "        if '.txt' == file_extension:  # ensure we only read from *.txt files\n",
        "            with open(file_path, encoding='utf8') as f_open:\n",
        "                content_lines = f_open.read()  # grab the contents of the file\n",
        "            content += content_lines\n",
        "    return content"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 72,
      "metadata": {
        "id": "nsZNguEYiAdR"
      },
      "outputs": [],
      "source": [
        "def clean_document(document: str) -> str:\n",
        "    \"\"\"\n",
        "        Perform text and character removal\n",
        "    \"\"\"\n",
        "    # Normalize text by converting to lower case to minimise query time, computational and memory costs\n",
        "    document = document.lower()\n",
        "    # Remove redundant characters directly with custom regex to ease tokenization\n",
        "    reg_to_drop = r'''(?x)           # flag to allow comments and multi-line regex\n",
        "              | \\( | \\)              # remove ()-brackets as their contents can be useful\n",
        "              | ^\\$[\\d,]+(\\.\\d*)?$   # cut out the money expressions\n",
        "              | [^(!\"#$%&'()*+, -.\\/:;<=>?@[\\]^_`{|}~) | (a-zA-Z0-9\\s)]  # remove non-word, non-digit, non-punctuation characters, non-space, e.g. '•', '≠'\n",
        "              | \\\\[u|U][a-zA-Z0-9]+  # remove codes like `\\u00F4`\n",
        "    '''\n",
        "    reg_compiler = re.compile(reg_to_drop)\n",
        "    document = reg_compiler.sub(r'', document)\n",
        "\n",
        "    # Handle apostrophes\n",
        "    document = re.sub(r\"won\\'t\", \"will not\", document)\n",
        "    document = re.sub(r\"n\\'t\", \" not\", document)\n",
        "    document = re.sub(r\"\\'t\", \" not\", document)\n",
        "    document = re.sub(r\"\\'m\", \" am\", document)\n",
        "    document = re.sub(r\"\\'re\", \" are\", document)\n",
        "    document = re.sub(r\"\\'ll\", \" will\", document)\n",
        "    document = re.sub(r\"\\'ve\", \" have\", document)\n",
        "\n",
        "    # Add extra spaces to \\n and \\#\n",
        "    document = re.sub(r\"\\n\", \"\\n \", document)\n",
        "    document = re.sub(r\"##\", \"## \", document)\n",
        "    document = re.sub(r\"\\.\\.\\.\", \" \", document)\n",
        "\n",
        "    # Handle shortenings\n",
        "    document = re.sub(r\" u \", \" you \", document)\n",
        "    # Transform sentiments\n",
        "    document = re.sub(r\"\\[\\+\\]\", \"[+1]\", document)\n",
        "    document = re.sub(r\"\\[1\\]\", \"[+1]\", document)\n",
        "    document = re.sub(r\"\\[u\\]\", \"\", document)\n",
        "    document = re.sub(r\"\\[p\\]\", \"\", document)\n",
        "    document = re.sub(r\"\\[s\\]\", \"\", document)\n",
        "    document = re.sub(r\"\\[cc\\]\", \"\", document)\n",
        "    document = re.sub(r\"\\[cs\\]\", \"\", document)\n",
        "\n",
        "    # drop any web links\n",
        "    document = re.sub(r\"([^ ]+ (?<=\\. [a-z] {3}))\", \"<url>\", document)\n",
        "\n",
        "    return document"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 73,
      "metadata": {
        "id": "C5Wp8wD1OF70"
      },
      "outputs": [],
      "source": [
        "def process_document(document: str, drop_punct: bool = True, drop_digits: bool = True):\n",
        "    \"\"\"\n",
        "      Pre-process a document and return a list of its terms\n",
        "    :param document:\n",
        "    :return:\n",
        "    \"\"\"\n",
        "    # Clean the data\n",
        "    document = clean_document(document)\n",
        "\n",
        "    # new lines are no longer needed as the review lines have been concatenated with a space\n",
        "    document = document.replace('\\n', '')\n",
        "\n",
        "    # Apply NLTK’s recommended word tokenizer for single word expressions (swe)\n",
        "    word_tokens = word_tokenize(document)\n",
        "\n",
        "    if drop_punct:\n",
        "        punc = r'''[\\!\\(\\)\\-\\[\\]\\{\\}\\;\\:\\'\\\"\\,\\<\\>\\.\\/\\?\\@\\$\\%\\\\\\^\\&\\*\\_\\~\\`]'''\n",
        "        word_tokens = [w for w in word_tokens if not re.search(punc, w)]\n",
        "        # word_tokens = [w for w in word_tokens if not w in ['.', ',', '#', 'l', '[', ']']]\n",
        "    \n",
        "    if drop_digits:\n",
        "        digit = r'''[0-9]+'''\n",
        "        word_tokens = [w for w in word_tokens if not re.search(digit, w)]\n",
        "\n",
        "    # Remove STOP words\n",
        "    stop_words = set(stopwords.words('english'))\n",
        "    filtered_data = [w for w in word_tokens if not w.lower() in stop_words]\n",
        "\n",
        "    stemmer = SnowballStemmer(language='english')\n",
        "    filtered_data = [stemmer.stem(w) for w in filtered_data]\n",
        "\n",
        "    return filtered_data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 74,
      "metadata": {
        "id": "YAmahE05noIH"
      },
      "outputs": [],
      "source": [
        "def parse_sentiment(sentiment_str: str, verbose: bool = False) -> int:\n",
        "    sentiment_score = 0\n",
        "    sentiment_regex = r'(?<=\\[).+?(?=\\])'\n",
        "    sentiment_scores_str = re.findall(sentiment_regex, sentiment_str)\n",
        "    if verbose:\n",
        "        print(sentiment_scores_str)\n",
        "    for s in sentiment_scores_str:\n",
        "        try:\n",
        "            sentiment_score += int(s)\n",
        "        except ValueError:\n",
        "            pass\n",
        "    return sentiment_score"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 75,
      "metadata": {
        "id": "LLIDDfoVo8GL"
      },
      "outputs": [],
      "source": [
        "assert parse_sentiment('router[+2]') == 2\n",
        "assert parse_sentiment('install[-2], router[+1]') == -1"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 76,
      "metadata": {
        "id": "0g-g6DN_j8U_"
      },
      "outputs": [],
      "source": [
        "def create_dataset(document: str):\n",
        "    \"\"\"\n",
        "        Create whole dataset with labels and sentiment score per review. \n",
        "        Also, concatenate all review lines together.\n",
        "    \"\"\"\n",
        "    dataset = []\n",
        "    # One of the files omits `[t]` hence there will be different assumptions on the authorship\n",
        "    REVIEW_MAX_CHAR_SIZE = 10_000  # hard-coded value based on distribution of review sizes\n",
        "    \n",
        "    # Iterate over each review `[t]`\n",
        "    for review in document.split('[t]'):\n",
        "        review_text = \"\"\n",
        "        sentiment = 0  # UNKNOWN\n",
        "        if len(review) <= REVIEW_MAX_CHAR_SIZE:\n",
        "            # Iterate over each review `\\n`\n",
        "            for line in review.split('\\n'):\n",
        "                # Split sentiment from review line\n",
        "                review_line_parts = line.split('##')\n",
        "                # no sentiment, only review line\n",
        "                if(len(review_line_parts) == 1):\n",
        "                    review_text += review_line_parts[0] + ' '\n",
        "                # sentiment and review line exist\n",
        "                elif(len(review_line_parts) > 1):\n",
        "                    sentiment += parse_sentiment(review_line_parts[0])\n",
        "                    review_text += review_line_parts[1] + ' '\n",
        "                else:\n",
        "                    print(review_line_parts)\n",
        "            # Define Positive, Neutral, Negative labels\n",
        "            label = 0\n",
        "            if sentiment > 0:\n",
        "                label = 1\n",
        "            elif sentiment < 0:\n",
        "                label = -1\n",
        "            review_entry = {\n",
        "                'label': label,\n",
        "                'sentiment_score': sentiment,\n",
        "                'text': review_text\n",
        "            }\n",
        "            dataset.append(review_entry)\n",
        "        else:  # since `[t]` is missing we will assume a review is a single line with a recorded sentiment\n",
        "            # Iterate over each review `\\n`\n",
        "            for line in review.split('\\n'):\n",
        "                review_text = \"\"\n",
        "                sentiment = 0  # UNKNOWN\n",
        "                review_line_parts = line.split('##')\n",
        "                # We only take the lines with sentiment and discard the rest for the faulty documents\n",
        "                if(len(review_line_parts) > 1):\n",
        "                    sentiment += parse_sentiment(review_line_parts[0])\n",
        "                    # Define Positive, Neutral, Negative labels\n",
        "                    review_text += review_line_parts[1]\n",
        "                    label = 0\n",
        "                    if sentiment > 0:\n",
        "                        label = 1\n",
        "                    elif sentiment < 0:\n",
        "                        label = -1\n",
        "                    review_entry = {\n",
        "                        'label': label,\n",
        "                        'sentiment_score': sentiment,\n",
        "                        'text': review_text\n",
        "                    }\n",
        "                    dataset.append(review_entry)\n",
        "    return dataset"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 77,
      "metadata": {
        "id": "7xwWoE9_gQVa"
      },
      "outputs": [],
      "source": [
        "content = read_data(drive_path)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 78,
      "metadata": {
        "id": "PUkjiMdQh9RF"
      },
      "outputs": [],
      "source": [
        "content = clean_document(content)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 79,
      "metadata": {
        "id": "2g-9IVdqruyf"
      },
      "outputs": [],
      "source": [
        "dataset = create_dataset(content)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 80,
      "metadata": {
        "id": "JPniPEi2s47l"
      },
      "outputs": [],
      "source": [
        "dataset = [entry for entry in dataset if entry['label']!=0]  # remove Neutral classes"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 81,
      "metadata": {
        "id": "arKm1AMqNrZO"
      },
      "outputs": [],
      "source": [
        "for entry in dataset:\n",
        "    entry['tokens'] = process_document(entry['text'])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JMgytkmnKF_0"
      },
      "source": [
        "#### Find stats about review sizes"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 82,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 424
        },
        "id": "ITt5gwi7KFJK",
        "outputId": "25138f3f-fa66-427b-bbf4-0620e36d7fec"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "     label  sentiment_score  \\\n",
              "0        1               19   \n",
              "1       -1               -7   \n",
              "2       -1               -3   \n",
              "3        1               24   \n",
              "4        1               12   \n",
              "..     ...              ...   \n",
              "582      1                3   \n",
              "583      1                2   \n",
              "584      1                2   \n",
              "585      1                2   \n",
              "586      1                2   \n",
              "\n",
              "                                                  text  \\\n",
              "0      this router does everything that it is suppo...   \n",
              "1      it's probably a fine router, but i found the...   \n",
              "2      if you are buying this product and operate o...   \n",
              "3      for an item that is not even available in be...   \n",
              "4      i had some issues when i had enabled wep enc...   \n",
              "..                                                 ...   \n",
              "582   the ear buds that come with it look cheap, bu...   \n",
              "583     the click wheel is incredibly convenient, h...   \n",
              "584    i am a big fan of itunes, simply because it ...   \n",
              "585                                 way to go apple!     \n",
              "586   i rarely get drawn in by mass-commercialism, ...   \n",
              "\n",
              "                                                tokens  tokens_size  \n",
              "0    [router, everyth, suppos, dont, realli, know, ...           86  \n",
              "1    [probabl, fine, router, found, instal, wizard,...          135  \n",
              "2    [buy, product, oper, mac, platform, ditch, cd,...          240  \n",
              "3    [item, even, avail, best, buy, major, store, w...          236  \n",
              "4    [issu, enabl, wep, encrypt, move, wpa, iron, l...           79  \n",
              "..                                                 ...          ...  \n",
              "582  [ear, bud, come, look, cheap, sound, qualiti, ...            8  \n",
              "583  [click, wheel, incred, conveni, howev, littl, ...           12  \n",
              "584  [big, fan, itun, simpli, terribl, fast, conven...           11  \n",
              "585                                    [way, go, appl]            3  \n",
              "586  [rare, get, drawn, someth, ipod, made, say, wo...           14  \n",
              "\n",
              "[587 rows x 5 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-e9321118-1637-4bd6-a230-7ff3d305d5a8\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <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>label</th>\n",
              "      <th>sentiment_score</th>\n",
              "      <th>text</th>\n",
              "      <th>tokens</th>\n",
              "      <th>tokens_size</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1</td>\n",
              "      <td>19</td>\n",
              "      <td>this router does everything that it is suppo...</td>\n",
              "      <td>[router, everyth, suppos, dont, realli, know, ...</td>\n",
              "      <td>86</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>-1</td>\n",
              "      <td>-7</td>\n",
              "      <td>it's probably a fine router, but i found the...</td>\n",
              "      <td>[probabl, fine, router, found, instal, wizard,...</td>\n",
              "      <td>135</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>-1</td>\n",
              "      <td>-3</td>\n",
              "      <td>if you are buying this product and operate o...</td>\n",
              "      <td>[buy, product, oper, mac, platform, ditch, cd,...</td>\n",
              "      <td>240</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>1</td>\n",
              "      <td>24</td>\n",
              "      <td>for an item that is not even available in be...</td>\n",
              "      <td>[item, even, avail, best, buy, major, store, w...</td>\n",
              "      <td>236</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1</td>\n",
              "      <td>12</td>\n",
              "      <td>i had some issues when i had enabled wep enc...</td>\n",
              "      <td>[issu, enabl, wep, encrypt, move, wpa, iron, l...</td>\n",
              "      <td>79</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>582</th>\n",
              "      <td>1</td>\n",
              "      <td>3</td>\n",
              "      <td>the ear buds that come with it look cheap, bu...</td>\n",
              "      <td>[ear, bud, come, look, cheap, sound, qualiti, ...</td>\n",
              "      <td>8</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>583</th>\n",
              "      <td>1</td>\n",
              "      <td>2</td>\n",
              "      <td>the click wheel is incredibly convenient, h...</td>\n",
              "      <td>[click, wheel, incred, conveni, howev, littl, ...</td>\n",
              "      <td>12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>584</th>\n",
              "      <td>1</td>\n",
              "      <td>2</td>\n",
              "      <td>i am a big fan of itunes, simply because it ...</td>\n",
              "      <td>[big, fan, itun, simpli, terribl, fast, conven...</td>\n",
              "      <td>11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>585</th>\n",
              "      <td>1</td>\n",
              "      <td>2</td>\n",
              "      <td>way to go apple!</td>\n",
              "      <td>[way, go, appl]</td>\n",
              "      <td>3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>586</th>\n",
              "      <td>1</td>\n",
              "      <td>2</td>\n",
              "      <td>i rarely get drawn in by mass-commercialism, ...</td>\n",
              "      <td>[rare, get, drawn, someth, ipod, made, say, wo...</td>\n",
              "      <td>14</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>587 rows × 5 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-e9321118-1637-4bd6-a230-7ff3d305d5a8')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-e9321118-1637-4bd6-a230-7ff3d305d5a8 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-e9321118-1637-4bd6-a230-7ff3d305d5a8');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 82
        }
      ],
      "source": [
        "dataset_df = pd.DataFrame(dataset)\n",
        "dataset_df['tokens_size'] = dataset_df['tokens'].apply(lambda x: len(x))\n",
        "dataset_df"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print('Positive/Negative review split:')\n",
        "dataset_df.value_counts('label')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pYBIB3y-X11f",
        "outputId": "367e0802-cece-44a4-b563-8efc7ea2b790"
      },
      "execution_count": 83,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Positive/Negative review split:\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "label\n",
              " 1    414\n",
              "-1    173\n",
              "dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 83
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vHMH73CPWpLq"
      },
      "source": [
        "#### Explore review/token statistics"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 84,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QFfOfwNJNXLq",
        "outputId": "63bd627c-64dd-4249-eceb-53c1c5fef2f3"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Biggest review has token size of: 564\n"
          ]
        }
      ],
      "source": [
        "print('Biggest review has token size of:', end=' ')\n",
        "s = dataset_df.tokens_size.max()\n",
        "print(s)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 85,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BWBVCcQJMcfG",
        "outputId": "ee87fe68-5d0b-4cf8-a678-d4daae8b633e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Statistically summarise the review word count: 587\n"
          ]
        }
      ],
      "source": [
        "print('Statistically summarise the review word count:', end=' ')\n",
        "max_word_count = int(dataset_df.tokens_size.describe().round(2).iloc[0])\n",
        "print(max_word_count)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xnQVMWauSuzs"
      },
      "source": [
        "#### Find out vocabulary size"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 86,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2cKqZv4rS1NY",
        "outputId": "b47b4a4b-a325-4387-9206-0fd966f3d326"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Vocabulary size of all tokens is 3641\n"
          ]
        }
      ],
      "source": [
        "def flatten(l):\n",
        "    return [item for sublist in l for item in sublist]\n",
        "\n",
        "vocab = set(flatten(list(dataset_df.tokens)))\n",
        "vocab_size = len(vocab)\n",
        "print(f\"Vocabulary size of all tokens is {vocab_size}\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jl59ULjprTpj"
      },
      "source": [
        "#### Index-Based Encoding: Unused in embedding for LSTM"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "J4eY1JzwrpZT"
      },
      "source": [
        "Another encoding that preserves the order of the words as they occur in the sentences is the Index-Based Encoding.\n",
        "\n",
        "Pros\n",
        "\n",
        "(+) preserves order\n",
        "\n",
        "(+) easy to implement \n",
        "\n",
        "Cons\n",
        "\n",
        "(-) introduces artificial distance between words\n",
        "\n",
        "(-) allows variable vector lengths <=> we need to apply padding"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "metadata": {
        "id": "GfYMFg3VrXAE"
      },
      "outputs": [],
      "source": [
        "def get_index_encoding(tokens, vocab) -> List[int]:\n",
        "    ind_enc = []\n",
        "    for t in tokens:\n",
        "        ind_enc.append(vocab.index(t) + 1)\n",
        "    return ind_enc\n",
        "\n",
        "\n",
        "def get_index_decoding(encoding, vocab) -> List[str]:\n",
        "    ind_dec = []\n",
        "    for t in encoding:\n",
        "        # we will encode padding as 0\n",
        "        if t >= 1:\n",
        "            ind_dec.append(vocab[t - 1])\n",
        "    return ind_dec\n",
        "\n",
        "\n",
        "assert get_index_encoding(['cat'], ['cat', 'dog', 'parrot']) == [1]\n",
        "assert get_index_decoding([0, 0, 1], ['cat', 'dog', 'parrot']) == ['cat']"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 26,
      "metadata": {
        "id": "rcEkoZQvtmas"
      },
      "outputs": [],
      "source": [
        "VOCAB_LIST = list(vocab)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 27,
      "metadata": {
        "id": "IW4s0dJftV4e"
      },
      "outputs": [],
      "source": [
        "dataset_df['tokens_encoded'] = dataset_df['tokens'].apply(lambda x: get_index_encoding(x, VOCAB_LIST))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KoxBwC_-voPJ"
      },
      "source": [
        "#### Split train and test data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 28,
      "metadata": {
        "id": "oy4aezEkvweC"
      },
      "outputs": [],
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "from keras.utils import pad_sequences"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 58,
      "metadata": {
        "id": "DVxhXcxHSGAq"
      },
      "outputs": [],
      "source": [
        "from keras.preprocessing.text import Tokenizer\n",
        "\n",
        "# Get cleaned (pre-processed) data\n",
        "list_sentences = dataset_df['text']\n",
        "\n",
        "max_features = vocab_size + 1\n",
        "maxlen = max_word_count + 1\n",
        "\n",
        "# Use LSTM default Tokenization\n",
        "tokenizer = Tokenizer(num_words=max_features)\n",
        "# Set vocabulary\n",
        "tokenizer.fit_on_texts(list(list_sentences))\n",
        "# Tokenize all text reviews\n",
        "list_tokenized_text = tokenizer.texts_to_sequences(list_sentences)\n",
        "# Pad all review sentences\n",
        "list_tokenized_text = pad_sequences(list_tokenized_text, maxlen=maxlen)\n",
        "\n",
        "# Transform into 2d numpy arrays\n",
        "x = np.array(list_tokenized_text)\n",
        "y = np.array(dataset_df['label'])\n",
        "\n",
        "# Ensure Negative is 0, Positive is 1 for binary cross-entropy \n",
        "y[y == -1] = 0\n",
        "y[y ==  1] = 1\n",
        "\n",
        "# Split data\n",
        "x_train, x_test, y_train, y_test = train_test_split(x, y)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eTLXsi61L899"
      },
      "source": [
        "#### Implement a Bi-directional LSTM"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 59,
      "metadata": {
        "id": "Vzu_my-yJhTV"
      },
      "outputs": [],
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras import layers\n",
        "from keras.layers import Embedding, LSTM, Dense, Dropout, Bidirectional, GlobalMaxPool1D\n",
        "from keras.models import Sequential"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 60,
      "metadata": {
        "id": "xv5noEttzi5q"
      },
      "outputs": [],
      "source": [
        "def biLSTM(lstm_dims: int = 128, \n",
        "           fully_connected_size: int = 50, \n",
        "           dropout: float = 0.1, \n",
        "           verbose: bool = False):\n",
        "    model = keras.Sequential([\n",
        "        Embedding(input_dim=max_features, output_dim=lstm_dims),\n",
        "        Bidirectional(layers.LSTM(lstm_dims, return_sequences=True)),\n",
        "        GlobalMaxPool1D(),\n",
        "        Dense(fully_connected_size,activation='relu'),\n",
        "        Dropout(dropout),\n",
        "        layers.Dense(1, activation=\"sigmoid\")\n",
        "    ])\n",
        "    if verbose:\n",
        "        print(model.summary())\n",
        "    return model  "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SLTEASuqzjoi"
      },
      "source": [
        "#### 5-fold Cross-validation"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 45,
      "metadata": {
        "id": "3Lt_2J8zzpcI"
      },
      "outputs": [],
      "source": [
        "from sklearn.model_selection import KFold"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ReObX4gfzp7A"
      },
      "outputs": [],
      "source": [
        "# hyperparams to test: lstm_dims, fully_connected_size, dropout\n",
        "hyperparameter_stats = []\n",
        "\n",
        "EPOCHS = 3\n",
        "num_folds = 5\n",
        "\n",
        "# the experiments are stored in different csv-s created by modifying filename below\n",
        "filename = 'hyperparameter_df_DEEP_LEARNING_TO_EPOCH_3_RENEWED.csv'\n",
        "csv_file_path = f'/content/drive/My Drive/Colab Notebooks/{filename}'\n",
        "\n",
        "# Set up the K-fold cross validation\n",
        "kfold = KFold(n_splits=num_folds, shuffle=True)\n",
        "\n",
        "for lstm_dims in [32, 64, 128, 180, 256]:\n",
        "    for fold, (train_index, test_index) in enumerate(kfold.split(x_train)):\n",
        "        start = time()\n",
        "\n",
        "        # create model\n",
        "        model = biLSTM(lstm_dims=lstm_dims)\n",
        "        model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "        # Fit the model on the train data\n",
        "        model.fit(x_train[train_index], y_train[train_index],\n",
        "                  epochs=EPOCHS,\n",
        "                  validation_data=(x_train[test_index], y_train[test_index]))\n",
        "        end = time()\n",
        "        \n",
        "        # Evaluate the model on the test data\n",
        "        _, acc = model.evaluate(x_test, y_test)\n",
        "        print(f'Fold {fold+1}: Test accuracy = {acc}')\n",
        "\n",
        "        # Record hyperparam tuning progress\n",
        "        hyperparameter_entry = {\n",
        "            'lstm_dims': lstm_dims,\n",
        "            'accuracy': np.round(acc, 2),\n",
        "            'epochs': EPOCHS,\n",
        "            'kfold': fold,\n",
        "            'time (s)': np.round(end-start, 2),\n",
        "            \n",
        "        }\n",
        "        print(hyperparameter_entry)\n",
        "        hyperparameter_stats.append(hyperparameter_entry)\n",
        "    hyperparameter_df = pd.DataFrame(hyperparameter_stats)\n",
        "    hyperparameter_df.to_csv(csv_file_path, index=False)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Selected model"
      ],
      "metadata": {
        "id": "pqDDRKds9nAz"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "lstm_dims = 32\n",
        "fully_connected_size = 50\n",
        "dropout = 0.1\n",
        "EPOCHS = 3\n",
        "\n",
        "kfold = KFold(n_splits=num_folds, shuffle=True)\n",
        "\n",
        "for fold, (train_index, test_index) in enumerate(kfold.split(x_train)):\n",
        "    # create model\n",
        "    model = biLSTM(lstm_dims=lstm_dims, \n",
        "                  fully_connected_size=fully_connected_size,\n",
        "                  dropout=dropout)\n",
        "    model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "    # Fit the model on the train data\n",
        "    model.fit(x_train[train_index], y_train[train_index],\n",
        "              epochs=EPOCHS,\n",
        "              validation_data=(x_train[test_index], y_train[test_index]))\n",
        "\n",
        "    # Evaluate the model on the test data\n",
        "    _, acc = model.evaluate(x_test, y_test)\n",
        "    print(f'Fold {fold+1}: Test accuracy = {acc}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9p99LgAu9qUO",
        "outputId": "41030071-236e-465e-d846-e0fce8325db6"
      },
      "execution_count": 61,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/3\n",
            "11/11 [==============================] - 9s 478ms/step - loss: 0.6813 - accuracy: 0.6619 - val_loss: 0.6603 - val_accuracy: 0.7273\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 363ms/step - loss: 0.6475 - accuracy: 0.6960 - val_loss: 0.6040 - val_accuracy: 0.7273\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 370ms/step - loss: 0.6253 - accuracy: 0.6960 - val_loss: 0.5862 - val_accuracy: 0.7273\n",
            "5/5 [==============================] - 0s 87ms/step - loss: 0.6001 - accuracy: 0.7143\n",
            "Fold 1: Test accuracy = 0.7142857313156128\n",
            "Epoch 1/3\n",
            "11/11 [==============================] - 9s 482ms/step - loss: 0.6710 - accuracy: 0.7102 - val_loss: 0.6538 - val_accuracy: 0.6705\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 377ms/step - loss: 0.6158 - accuracy: 0.7102 - val_loss: 0.6484 - val_accuracy: 0.6705\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 383ms/step - loss: 0.6059 - accuracy: 0.7102 - val_loss: 0.6365 - val_accuracy: 0.6705\n",
            "5/5 [==============================] - 0s 85ms/step - loss: 0.6002 - accuracy: 0.7143\n",
            "Fold 2: Test accuracy = 0.7142857313156128\n",
            "Epoch 1/3\n",
            "11/11 [==============================] - 8s 471ms/step - loss: 0.6820 - accuracy: 0.6534 - val_loss: 0.6584 - val_accuracy: 0.7500\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 370ms/step - loss: 0.6491 - accuracy: 0.6903 - val_loss: 0.5911 - val_accuracy: 0.7500\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 369ms/step - loss: 0.6233 - accuracy: 0.6903 - val_loss: 0.5625 - val_accuracy: 0.7500\n",
            "5/5 [==============================] - 0s 84ms/step - loss: 0.5995 - accuracy: 0.7143\n",
            "Fold 3: Test accuracy = 0.7142857313156128\n",
            "Epoch 1/3\n",
            "11/11 [==============================] - 9s 479ms/step - loss: 0.6713 - accuracy: 0.7045 - val_loss: 0.6502 - val_accuracy: 0.6818\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 373ms/step - loss: 0.6218 - accuracy: 0.7074 - val_loss: 0.6218 - val_accuracy: 0.6818\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 379ms/step - loss: 0.6039 - accuracy: 0.7074 - val_loss: 0.6214 - val_accuracy: 0.6818\n",
            "5/5 [==============================] - 0s 87ms/step - loss: 0.5984 - accuracy: 0.7143\n",
            "Fold 4: Test accuracy = 0.7142857313156128\n",
            "Epoch 1/3\n",
            "11/11 [==============================] - 8s 474ms/step - loss: 0.6760 - accuracy: 0.6733 - val_loss: 0.6589 - val_accuracy: 0.6818\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 373ms/step - loss: 0.6197 - accuracy: 0.7074 - val_loss: 0.6330 - val_accuracy: 0.6818\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 385ms/step - loss: 0.6034 - accuracy: 0.7074 - val_loss: 0.6377 - val_accuracy: 0.6818\n",
            "5/5 [==============================] - 0s 86ms/step - loss: 0.6024 - accuracy: 0.7143\n",
            "Fold 5: Test accuracy = 0.7142857313156128\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Selected model with stratified data"
      ],
      "metadata": {
        "id": "GPEpfG8Th5o0"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.model_selection import StratifiedKFold"
      ],
      "metadata": {
        "id": "lobur-CQiXB_"
      },
      "execution_count": 62,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "lstm_dims = 32\n",
        "fully_connected_size = 10\n",
        "dropout = 0.1\n",
        "EPOCHS = 3\n",
        "\n",
        "kfold = StratifiedKFold(n_splits=num_folds, shuffle=True)\n",
        "\n",
        "for fold, (train_index, test_index) in enumerate(kfold.split(x_train, y_train)):\n",
        "    # create model\n",
        "    model = biLSTM(lstm_dims=lstm_dims, \n",
        "                   fully_connected_size=fully_connected_size,\n",
        "                   dropout=dropout)\n",
        "    model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "    # Fit the model on the train data\n",
        "    model.fit(x[train_index], y[train_index],\n",
        "              epochs=EPOCHS,\n",
        "              validation_data=(x[test_index], y[test_index]))\n",
        "    \n",
        "    # Evaluate the model on the test data\n",
        "    _, acc = model.evaluate(x_test, y_test)\n",
        "    print(f'Fold {fold+1}: Test accuracy = {acc}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lTGBW7TEh-UQ",
        "outputId": "b473de96-f6ba-4ebd-a8b8-3f581737db8d"
      },
      "execution_count": 64,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/3\n",
            "11/11 [==============================] - 8s 474ms/step - loss: 0.6788 - accuracy: 0.6989 - val_loss: 0.6562 - val_accuracy: 0.7386\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 372ms/step - loss: 0.6430 - accuracy: 0.7102 - val_loss: 0.5879 - val_accuracy: 0.7386\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 367ms/step - loss: 0.6350 - accuracy: 0.7102 - val_loss: 0.5748 - val_accuracy: 0.7386\n",
            "5/5 [==============================] - 0s 83ms/step - loss: 0.5985 - accuracy: 0.7143\n",
            "Fold 1: Test accuracy = 0.7142857313156128\n",
            "Epoch 1/3\n",
            "11/11 [==============================] - 48s 486ms/step - loss: 0.6858 - accuracy: 0.6080 - val_loss: 0.6748 - val_accuracy: 0.6932\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 6s 554ms/step - loss: 0.6548 - accuracy: 0.7216 - val_loss: 0.6390 - val_accuracy: 0.6932\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 382ms/step - loss: 0.6075 - accuracy: 0.7216 - val_loss: 0.6230 - val_accuracy: 0.6932\n",
            "5/5 [==============================] - 0s 83ms/step - loss: 0.5999 - accuracy: 0.7143\n",
            "Fold 2: Test accuracy = 0.7142857313156128\n",
            "Epoch 1/3\n",
            "11/11 [==============================] - 9s 474ms/step - loss: 0.6753 - accuracy: 0.6989 - val_loss: 0.6362 - val_accuracy: 0.7727\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 370ms/step - loss: 0.6291 - accuracy: 0.7017 - val_loss: 0.5653 - val_accuracy: 0.7727\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 372ms/step - loss: 0.6099 - accuracy: 0.7017 - val_loss: 0.5433 - val_accuracy: 0.7727\n",
            "5/5 [==============================] - 0s 85ms/step - loss: 0.5962 - accuracy: 0.7143\n",
            "Fold 3: Test accuracy = 0.7142857313156128\n",
            "Epoch 1/3\n",
            "11/11 [==============================] - 9s 471ms/step - loss: 0.6859 - accuracy: 0.7017 - val_loss: 0.6710 - val_accuracy: 0.7500\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 376ms/step - loss: 0.6586 - accuracy: 0.7074 - val_loss: 0.6170 - val_accuracy: 0.7500\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 370ms/step - loss: 0.6059 - accuracy: 0.7074 - val_loss: 0.5647 - val_accuracy: 0.7500\n",
            "5/5 [==============================] - 0s 82ms/step - loss: 0.5986 - accuracy: 0.7143\n",
            "Fold 4: Test accuracy = 0.7142857313156128\n",
            "Epoch 1/3\n",
            "11/11 [==============================] - 9s 478ms/step - loss: 0.6744 - accuracy: 0.7301 - val_loss: 0.6678 - val_accuracy: 0.6250\n",
            "Epoch 2/3\n",
            "11/11 [==============================] - 4s 375ms/step - loss: 0.6045 - accuracy: 0.7386 - val_loss: 0.6770 - val_accuracy: 0.6250\n",
            "Epoch 3/3\n",
            "11/11 [==============================] - 4s 375ms/step - loss: 0.5811 - accuracy: 0.7386 - val_loss: 0.7159 - val_accuracy: 0.6250\n",
            "5/5 [==============================] - 0s 83ms/step - loss: 0.6059 - accuracy: 0.7143\n",
            "Fold 5: Test accuracy = 0.7142857313156128\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7afEnxLgannk"
      },
      "source": [
        "### Analyse EPOCH = 10 effect on Hyperparameters"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sXMpsLTHe4uu"
      },
      "source": [
        "Observations from output file `hyperparameter_to_epoch_10.txt` suggest that **EPOCHS > 3** result in **increased training accuracy**, but **unchanged validation accuracy** (I suspect this is due to over-fitting to the training data). Hence, for other experiments I will assume MAX EPOCHS = 3 to maintain a quicker testing environment. Additionally, Dropout did not seem to affect considerably the accuracy."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 65,
      "metadata": {
        "id": "X6Y7TceFas0-"
      },
      "outputs": [],
      "source": [
        "# file will be zipped with submission\n",
        "hyperparameter_df = pd.read_csv(f'/content/drive/My Drive/Colab Notebooks/hyperparameter_df_DEEP_LEARNING_TO_EPOCH_10.csv')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U8jWRHNRcw_n"
      },
      "source": [
        "#### Show dropout rate change"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "ZZKh0IiDZBNj",
        "outputId": "8b81053b-ead5-43e8-b38b-fad0e821530b"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.title('No significant difference between different Dropout rates')\n",
        "sns.boxplot(data=hyperparameter_df, x='dropout', y='accuracy', hue='lstm_dims')\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.title('No significant difference between different fully_connected_sizes')\n",
        "sns.boxplot(data=hyperparameter_df, x='fully_connected_size', y='accuracy', hue='lstm_dims')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 296
        },
        "id": "acsd2DQw31c3",
        "outputId": "e8a4ee3d-f7c0-43e9-eaaf-7c9b9c8311ba"
      },
      "execution_count": 67,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Analyse EPOCH = 3 effect on Hyperparameters"
      ],
      "metadata": {
        "id": "cBSB6st_8Zcz"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Experimentation with different LSTM sizes at Epoch 3 **does not display any differenes wrt accuracy** between selected dimensions [32, 64, 128, 180, 256]"
      ],
      "metadata": {
        "id": "qoD1JqvJ8yd_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# file will be zipped with submission\n",
        "hyperparameter_df = pd.read_csv(f'/content/drive/My Drive/Colab Notebooks/hyperparameter_df_DEEP_LEARNING_TO_EPOCH_3.csv')"
      ],
      "metadata": {
        "id": "Cirx6G9n8fMr"
      },
      "execution_count": 69,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plt.title('No difference between LSTM dimensions')\n",
        "sns.barplot(data=hyperparameter_df, x='lstm_dims', y='accuracy')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 296
        },
        "id": "B67O6FUH8gNI",
        "outputId": "33c4e2c9-e9a2-4d6d-eb47-e4694799e93e"
      },
      "execution_count": 70,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "-bh8ScLa8g0b"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}