11_multiclass_classification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# default_exp multiclass_classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#hide\n",
    "from nbdev.showdoc import *\n",
    "# all_flag"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Muticlass Classification on Imbalanced Dataset\n",
    "\n",
    "> Summary: Multiclass Classification, Naive Bayes, Logistic Regression, SVM, Random Forest, XGBoosting, BERT, Imbalanced Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Task:** The goal of this project is to build a classification model to accurately classify text documents into a predefined category. The dataset consists of a collection of customer complaints in the form of free text along with their corresponding departments (i.e. predifined categories). The original dataset is available [here](https://www.consumerfinance.gov/data-research/consumer-complaints/). However, I created a new dataset from it, which only includes \"complaint id\", \"product_group\" and \"text\" columns.\n",
    "\n",
    "In order to achieve this goal, I performed the following steps:\n",
    "1. Get/Load the dataset\n",
    "\n",
    "2. Explore the data\n",
    "\n",
    "3. Prepare the data\n",
    "\n",
    "4. Build the model\n",
    "\n",
    "5. Fine-tune the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "import os\n",
    "import time\n",
    "import re\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "\n",
    "from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer\n",
    "from sklearn import preprocessing\n",
    "from sklearn.model_selection import train_test_split, StratifiedKFold, KFold\n",
    "\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.svm import LinearSVC\n",
    "from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier\n",
    "\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import plot_confusion_matrix\n",
    "from sklearn.metrics import precision_recall_fscore_support as score"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "def load_data(file_path):\n",
    "    \"\"\" Load the csv file and return a Dataframe. \"\"\"\n",
    "    df = pd.read_csv(file_path, delimiter=',')\n",
    "    return df\n",
    "\n",
    "def clean_text(text):\n",
    "    \"\"\" clean the text by removing special characters, punctuations, etc. \"\"\"\n",
    "    text = text.lower()\n",
    "    remove_xx_seq  = re.compile('xx+')\n",
    "    replace_by_space = re.compile('[/(){}\\[\\]\\|@,;]')\n",
    "    remove_special_characters = re.compile('[^0-9a-z +]')\n",
    "    remove_extra_space = re.compile('[^A-Za-z0-9]+')\n",
    "    remove_numbers = re.compile('[0-9]+')\n",
    "    text = re.sub(remove_xx_seq, ' ', text)\n",
    "    text = re.sub(replace_by_space, ' ', text)\n",
    "    text = re.sub(remove_special_characters, ' ', text)\n",
    "    text = re.sub(remove_numbers, ' ', text)\n",
    "    text = re.sub(remove_extra_space, ' ', text)\n",
    "    return text.strip()\n",
    "\n",
    "\n",
    "def preprocess_corpus(df, column='text'):\n",
    "    \"\"\" Preprocess the entire corpus including cleaning the text documents and return the updated dataframe. \"\"\"\n",
    "    df[column] = df[column].apply(clean_text)\n",
    "    return df\n",
    "\n",
    "def encode_labels(labels):\n",
    "    \"\"\" Encode the class labels into a numbers. \"\"\"\n",
    "    label_encoder = preprocessing.LabelEncoder()\n",
    "    label_encoder.fit(labels)\n",
    "    return label_encoder\n",
    "\n",
    "def compute_tfidf(corpus, stop_words='english', ngram_range=(1, 1), max_features=None):\n",
    "    \"\"\" Calculate the tfidf features for all the text documents and return a (documents, fatures) matrix. \"\"\"\n",
    "    vectorizer = TfidfVectorizer(input='content', stop_words=stop_words, ngram_range=ngram_range, min_df=3, max_df=0.9, max_features=max_features)\n",
    "    print('Computing tfidf features...', end='')\n",
    "    X = vectorizer.fit_transform(corpus)\n",
    "    print('done!')\n",
    "    return X, vectorizer\n",
    "\n",
    "\n",
    "def train_test_model(model, X_train, X_test, y_train, y_test,labels):\n",
    "    \"\"\" Train and test the model using the training and test data sets. Return the predictions, accuracy and metric reports. \"\"\"\n",
    "    print('Start training...', end='')\n",
    "    model.fit(X_train, y_train)\n",
    "    print('done!')\n",
    "    print('Start testing...', end='')\n",
    "    predictions = model.predict(X_test)\n",
    "    accuracy = model.score(X_test, y_test)\n",
    "    metrics_report = classification_report(y_test, predictions, target_names=labels)\n",
    "    precision, recall, fscore, train_support = score(y_test, predictions, average='weighted')\n",
    "    print('done!')\n",
    "    return predictions, accuracy, metrics_report, (precision, recall, fscore) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Load the data...done.\n"
     ]
    }
   ],
   "source": [
    "file_path = os.path.join('wells_data', 'case_study_data.csv')\n",
    "print('Load the data...', end='')\n",
    "df = load_data(file_path)\n",
    "print('done.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Explore the Data\n",
    "\n",
    "After the data is loaded, we should examine and gain insight from the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>complaint_id</th>\n",
       "      <th>product_group</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2815595</td>\n",
       "      <td>bank_service</td>\n",
       "      <td>On XX/XX/2017 my check # XXXX was debited from...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2217937</td>\n",
       "      <td>bank_service</td>\n",
       "      <td>I opened a Bank of the the West account. The a...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2657456</td>\n",
       "      <td>bank_service</td>\n",
       "      <td>wells fargo in nj opened a business account wi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1414106</td>\n",
       "      <td>bank_service</td>\n",
       "      <td>A hold was placed on my saving account ( XXXX ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1999158</td>\n",
       "      <td>bank_service</td>\n",
       "      <td>Dear CFPB : I need to send a major concern/com...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   complaint_id product_group  \\\n",
       "0       2815595  bank_service   \n",
       "1       2217937  bank_service   \n",
       "2       2657456  bank_service   \n",
       "3       1414106  bank_service   \n",
       "4       1999158  bank_service   \n",
       "\n",
       "                                                text  \n",
       "0  On XX/XX/2017 my check # XXXX was debited from...  \n",
       "1  I opened a Bank of the the West account. The a...  \n",
       "2  wells fargo in nj opened a business account wi...  \n",
       "3  A hold was placed on my saving account ( XXXX ...  \n",
       "4  Dear CFPB : I need to send a major concern/com...  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of categories:  ['bank_service' 'credit_card' 'credit_reporting' 'debt_collection' 'loan'\n",
      " 'money_transfers' 'mortgage']\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7ff01a860b50>"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "categories = df.product_group.unique()\n",
    "print('Number of categories: ', categories)\n",
    "print()\n",
    "df.product_group.value_counts().plot(kind='bar', title='Categories vs Number of Documents', cmap='plasma')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The bar chart above shows that our dataset is imbalanced, i.e. the number of observations per class is not equally distributed. For example, the number of documents belonging to \"credit_reporting\" is more than 8 times of \"money_transfers\" documents. There are several common ways to deal with imbalanced datasets. However, I decided to use a method called *stratification* or *stratified sampling*. This technique removes the proportions variance, meaning it maintains the same class distribution in each subset of the complete training dataset while training, which gives more stability during the training. Particularly, I use *stratified 10-fold cross-validation* and *stratified version of a train/test split* approaches to managing this imbalanced dataset. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create a dataframe that contain number of words for each document \n",
    "dlength_df = pd.DataFrame({'doc_length': df.text.apply(lambda x: len(x.split()))})\n",
    "\n",
    "# Group the documents based on their number of words (i.e. length)\n",
    "grouped = dlength_df.groupby('doc_length')\n",
    "\n",
    "indices = grouped.indices\n",
    "word_count = []\n",
    "doc_count = []\n",
    "counter = 0\n",
    "for w,d in indices.items():\n",
    "    word_count.append(w)\n",
    "    doc_count.append(len(d))\n",
    "\n",
    "# Plot the distribution of words vs documents in the corpus\n",
    "plt.figure(figsize=(8,5))\n",
    "plt.plot(word_count, doc_count)\n",
    "plt.xlabel('Word count in document')\n",
    "plt.ylabel('Number of Documents')\n",
    "plt.title('Word count vs Number of documents')\n",
    "\n",
    "# Plot the Cumulative distribution of documents length\n",
    "plt.figure(figsize=(8,5))\n",
    "plt.hist(word_count, 30, density=True, histtype='step', cumulative=True, label='Complaints', color='red', linewidth=2)\n",
    "plt.xlabel('Word count in document')\n",
    "plt.ylabel('Fraction of Documents')\n",
    "plt.legend(loc='upper left')\n",
    "plt.title('Cumulative distribution of documents length')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By looking at the charts above, we realize that more than 70% of the documents have less than 1000 words. Also, the fraction of documents with 2000 words and more is less than 5%."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocess the Dataset\n",
    "\n",
    "One of the primary tasks before developing a model on textual data is cleaning the text. It usually includes lowercasing the text, removing stop words, removing puncutations, alphanumeric and/or special characters and so on. The functions `preprocess_corpus()` and `clean_text()` will clean up our dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "I 've tried disputing a couple of fraudulent accounts on my credit report, and the Bureau has taken some of it off but only the collection things. I 've sent letters to this company, XXXX XXXX, and XXXX XXXX  and have gotten little response. Not sure what is being looked into when the bureaus send me the results showing verified. How can they verify something that 's not mine? Most recently about a month ago I did send in an Identity Theft document and all of the items it asked me to include. Feel like after six months of disputing and working hard that it 's really going nowhere. Would like to have this stuff taken off my credit report bc it is not mine. It has been well over 30 days on more than one occasion, and neither XXXX XXXX or this bureau has done anything to address the creditors disregard for the federal laws. I leave out of the country for months at a time, and someone must have used my information and apply for credits because I was always getting declined no one could tell me what happened all I know is I never had XXXX XXXX account. I found out that the bureaus collect money from creditors when it comes to credit reporting consumers. Read that the agencies make more off bad reporting than good which makes me think this is being kept on my credit report so this company can profit off my misfortune.\n",
      "\n",
      "I opened up a Coinbase account to trade Cryptocurrencies. First purchase on XX/XX/XXXX. XXXX bitcoin for XXXX. Second purchase on XX/XX/XXXX. XXXX Etherium for XXXX dollars. Sold XXXX bitcoin for XXXX ( after fees ) on XX/XX/XXXX. Sold XXXX etherium for XXXX on XX/XX/XXXX. Currently have {$180.00} in a stagnant USD account on coinbase. I am unable to withdrawal the money to either a bank account or credit/debit card. I have not taken action with the company, because upon research this is not an uncommon problem and the company does not take action when it occurs. Per my research this is the best coarse of action to get my money back.\n",
      "\n",
      "I had a repossession back in early XX/XX/XXXX with a balance of 2100.00. I purchased a vehicle for my sons father and he did n't pay when we split. In XX/XX/XXXX, I disputed the debt after 7 years and the repo was taken off my credit report. Today I received a letter insisting I pay the debt. There are Barr laws in NC. I have n't heard from this debtor in years since I had the debt removed from my credit. It 's been 10 years.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Randomly show 3 complaints\n",
    "samples = [np.random.randint(len(df)) for i in range(3)]\n",
    "for index in samples:\n",
    "    print(df['text'][index])\n",
    "    print()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cleaning the text...done.\n"
     ]
    }
   ],
   "source": [
    "# Preprocess the corpus, i.e. clean the text\n",
    "print('Cleaning the text...', end='')\n",
    "df = preprocess_corpus(df, column='text')\n",
    "print('done.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare the Dataset\n",
    "\n",
    "After text preprocessing, we need to perform feature engineering and data prepration. We convert the text to vector representation using tfidf method. In this process, I consider the following options:\n",
    "\n",
    "- Remove stop words\n",
    "\n",
    "- Remove domain specific stop words, i.e. words that are rare (appear in 3 documents or less) or too common (occur in more that 90% of the documents)\n",
    "\n",
    "- Utilize unigrams, bigrams, etc\n",
    "\n",
    "- Use all the features (words) or a subset of them\n",
    "\n",
    "- Encode the class labels into numbers\n",
    "\n",
    "Please see the function `compute_tfidf()` for more information."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing tfidf features...done!\n"
     ]
    }
   ],
   "source": [
    "# Create tfidf features from the text\n",
    "stop_words    = 'english'\n",
    "ngram_range   = (1, 1)\n",
    "max_features  = None\n",
    "X, vectorizer = compute_tfidf(df['text'], stop_words, ngram_range, max_features)\n",
    "\n",
    "# Encode the labels\n",
    "labels = df.product_group.unique()\n",
    "label_encoder = encode_labels(labels)\n",
    "y = label_encoder.transform(df.product_group)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Split the Dataset into Train and Test Sets\n",
    "\n",
    "We make use of stratified `train_test_split()` function. This function will determine the distributions of classes and maintain the same distribution for both train and test sets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set Shape: (214688, 33028)  | Test set Shape: (53673, 33028)\n"
     ]
    }
   ],
   "source": [
    "# Split the dataset\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, stratify=y)\n",
    "print('Training set Shape: {}  | Test set Shape: {}'.format(X_train.shape, X_test.shape))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build the Model\n",
    "\n",
    "In this section, I will several different models for our multiclass classification task and compare them at the end. There are a wide variety of techniques that can be used. However, I decided to apply the following methods:\n",
    "\n",
    "- Naive Bayes (NB)\n",
    "\n",
    "- Logistic Regression\n",
    "\n",
    "- Linear SVM\n",
    "\n",
    "- Random Forest\n",
    "\n",
    "- XGBoosting\n",
    "\n",
    "- BERT"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Naive Bayes Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of documents = 214688  |  Number of features = 33028\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Total time: 0.28s\n",
      "accuracy: 0.809028748160155\n",
      "====================================================================================================\n",
      "                  precision    recall  f1-score   support\n",
      "\n",
      "    bank_service       0.79      0.70      0.74      4014\n",
      "     credit_card       0.75      0.75      0.75      5911\n",
      "credit_reporting       0.79      0.88      0.83     16246\n",
      " debt_collection       0.81      0.79      0.80     12292\n",
      "            loan       0.82      0.71      0.76      6207\n",
      " money_transfers       0.98      0.30      0.46       947\n",
      "        mortgage       0.88      0.93      0.90      8056\n",
      "\n",
      "        accuracy                           0.81     53673\n",
      "       macro avg       0.83      0.72      0.75     53673\n",
      "    weighted avg       0.81      0.81      0.81     53673\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7ff128c65d50>"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nb_model = MultinomialNB()\n",
    "print('Number of documents = {}  |  Number of features = {}'.format(X_train.shape[0], X_train.shape[1]))\n",
    "st_time = time.time()\n",
    "predictions, accuracy, metrics_report, nb_prf = train_test_model(nb_model, X_train, X_test, y_train, y_test, labels)\n",
    "en_time = time.time()\n",
    "print('Total time: {:.2f}s'.format(en_time-st_time))\n",
    "print('accuracy: {}'.format(accuracy))\n",
    "print('='*100)\n",
    "print(metrics_report)\n",
    "plot_confusion_matrix(nb_model, X_test, y_test, display_labels=labels, xticks_rotation='vertical', cmap=\"BuPu\")\n",
    "# plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of documents = 214688  |  Number of features = 33028\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Total time: 111.02s\n",
      "accuracy: 0.84783783280234\n",
      "====================================================================================================\n",
      "                  precision    recall  f1-score   support\n",
      "\n",
      "    bank_service       0.81      0.81      0.81      4014\n",
      "     credit_card       0.81      0.81      0.81      5911\n",
      "credit_reporting       0.86      0.86      0.86     16246\n",
      " debt_collection       0.83      0.84      0.84     12292\n",
      "            loan       0.82      0.77      0.80      6207\n",
      " money_transfers       0.83      0.73      0.78       947\n",
      "        mortgage       0.92      0.93      0.93      8056\n",
      "\n",
      "        accuracy                           0.85     53673\n",
      "       macro avg       0.84      0.82      0.83     53673\n",
      "    weighted avg       0.85      0.85      0.85     53673\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7ff0181d5ed0>"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "log_model = LogisticRegression(penalty='l2', max_iter=500)\n",
    "print('Number of documents = {}  |  Number of features = {}'.format(X_train.shape[0], X_train.shape[1]))\n",
    "st_time = time.time()\n",
    "predictions, accuracy, metrics_report, log_prf = train_test_model(log_model, X_train, X_test, y_train, y_test, labels)\n",
    "en_time = time.time()\n",
    "print('Total time: {:.2f}s'.format(en_time-st_time))\n",
    "print('accuracy: {}'.format(accuracy))\n",
    "print('='*100)\n",
    "print(metrics_report)\n",
    "plot_confusion_matrix(log_model, X_test, y_test, display_labels=labels, xticks_rotation='vertical', cmap=\"BuPu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of documents = 214688  |  Number of features = 33028\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Total time: 5.59s\n",
      "accuracy: 0.8447822927729026\n",
      "====================================================================================================\n",
      "                  precision    recall  f1-score   support\n",
      "\n",
      "    bank_service       0.77      0.82      0.80      4014\n",
      "     credit_card       0.78      0.84      0.81      5911\n",
      "credit_reporting       0.89      0.82      0.86     16246\n",
      " debt_collection       0.84      0.84      0.84     12292\n",
      "            loan       0.79      0.81      0.80      6207\n",
      " money_transfers       0.69      0.82      0.75       947\n",
      "        mortgage       0.91      0.94      0.93      8056\n",
      "\n",
      "        accuracy                           0.84     53673\n",
      "       macro avg       0.81      0.84      0.83     53673\n",
      "    weighted avg       0.85      0.84      0.85     53673\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7ff01a860bd0>"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "svm = LinearSVC(class_weight='balanced', verbose=False, max_iter=10000, tol=1e-4, C=0.1)\n",
    "print('Number of documents = {}  |  Number of features = {}'.format(X_train.shape[0], X_train.shape[1]))\n",
    "st_time = time.time()\n",
    "predictions, accuracy, metrics_report, svm_prf = train_test_model(svm, X_train, X_test, y_train, y_test, labels)\n",
    "en_time = time.time()\n",
    "print('Total time: {:.2f}s'.format(en_time-st_time))\n",
    "print('accuracy: {}'.format(accuracy))\n",
    "print('='*100)\n",
    "print(metrics_report)\n",
    "plot_confusion_matrix(svm, X_test, y_test, display_labels=labels, xticks_rotation='vertical', cmap=\"BuPu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of documents = 214688  |  Number of features = 33028\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Total time: 280.10s\n",
      "accuracy: 0.8414659139604643\n",
      "====================================================================================================\n",
      "                  precision    recall  f1-score   support\n",
      "\n",
      "    bank_service       0.81      0.76      0.79      4014\n",
      "     credit_card       0.82      0.75      0.79      5911\n",
      "credit_reporting       0.82      0.93      0.87     16246\n",
      " debt_collection       0.84      0.84      0.84     12292\n",
      "            loan       0.88      0.69      0.77      6207\n",
      " money_transfers       0.93      0.44      0.60       947\n",
      "        mortgage       0.90      0.93      0.91      8056\n",
      "\n",
      "        accuracy                           0.84     53673\n",
      "       macro avg       0.86      0.76      0.80     53673\n",
      "    weighted avg       0.84      0.84      0.84     53673\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7ff01a899c50>"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "rf_model = RandomForestClassifier(n_estimators=100, max_depth=100, min_samples_split=10, n_jobs=-1, verbose=0)\n",
    "print('Number of documents = {}  |  Number of features = {}'.format(X_train.shape[0], X_train.shape[1]))\n",
    "st_time = time.time()\n",
    "predictions, accuracy, metrics_report, rf_prf = train_test_model(rf_model, X_train, X_test, y_train, y_test, labels)\n",
    "en_time = time.time()\n",
    "print('Total time: {:.2f}s'.format(en_time-st_time))\n",
    "print('accuracy: {}'.format(accuracy))\n",
    "print('='*100)\n",
    "print(metrics_report)\n",
    "plot_confusion_matrix(rf_model, X_test, y_test, display_labels=labels, xticks_rotation='vertical', cmap=\"BuPu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Gradient Boosting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "gb_model = GradientBoostingClassifier(n_estimators=50, max_depth=10)\n",
    "predictions, accuracy, metrics_report, gb_prf = train_test_model(gb_model, X_train, X_test, y_train, y_test, labels)\n",
    "print('accuracy: {}'.format(accuracy))\n",
    "print(metrics_report)\n",
    "plot_confusion_matrix(gb_model, X_test, y_test, display_labels=labels, xticks_rotation='vertical', cmap=\"BuPu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```\n",
    "\n",
    "Start training...\n",
    "      Iter       Train Loss   Remaining Time \n",
    "         1      274073.2086          104.34m\n",
    "         2      240114.7651          104.04m\n",
    "         3      215548.5383          102.49m\n",
    "         4      196626.8849          100.89m\n",
    "         5      181335.8100           99.41m\n",
    "         6      168818.3973           97.47m\n",
    "         7      158202.7714           95.42m\n",
    "         8      149309.3867           93.23m\n",
    "         9      141564.8736           91.14m\n",
    "        10      134992.9559           88.86m\n",
    "        20       97935.1018           65.44m\n",
    "        30       82408.8805           42.23m\n",
    "        40       73946.2785           20.42m\n",
    "        50       67992.7312            0.00s\n",
    "Training done!\n",
    "Number of documents = 214688  |  Number of features = 10000\n",
    "\n",
    "Start testing...\n",
    "done!\n",
    "accuracy: 0.8323179252137947\n",
    "                  precision    recall  f1-score   support\n",
    "\n",
    "    bank_service       0.80      0.78      0.79      3994\n",
    "     credit_card       0.80      0.78      0.79      5964\n",
    "credit_reporting       0.85      0.86      0.86     16310\n",
    " debt_collection       0.79      0.83      0.81     12229\n",
    "            loan       0.80      0.77      0.79      6139\n",
    " money_transfers       0.75      0.63      0.68       911\n",
    "        mortgage       0.92      0.90      0.91      8126\n",
    "\n",
    "        accuracy                           0.83     53673\n",
    "       macro avg       0.82      0.79      0.80     53673\n",
    "    weighted avg       0.83      0.83      0.83     53673\n",
    "\n",
    "\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Issues with Gradient Boosting\n",
    "\n",
    "In gradient boosting trees are built sequentially, therefore, training takes longer time. In our case, since the dataset is large it even makes the training take significantly longer. As you can observe in the above cell, training took over 104 minutes and the accuracy is 83%, which ranks it at fourth positions only slightly better than Naive Bayes classifier.\n",
    "\n",
    "Since I was willing to try a few other configuration of the gradient boosting and see if it improves the performance, I decided to install the GPU supported version of XGBoosting algorithm on Google Colab. After searching, I found the XGBoosting library from [here](https://s3-us-west-2.amazonaws.com/xgboost-wheels/list.html) and installed it on Google Colab. After that, I implemented two forms of XGBoosting algorithm:\n",
    "\n",
    "1. **xgb:** This is the direct xgboost library.\n",
    "\n",
    "2. **XGBClassifier:** This is a sklearn wrapper for XGBoost. This enables us to use sklearn’s Grid Search with parallel processing.\n",
    "\n",
    "The next two cells includes two versions of XGBoost algorithm I implemented and copied the code from Colab. Although the gpu-supported version were much faster, the performance didn't improve."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### XGBClassifier Implementation for sklearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import xgboost as xgb\n",
    "from xgboost.sklearn import XGBClassifier\n",
    "from sklearn import metrics\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "xgb_model = XGBClassifier(objective='multi:softmax', n_estimators=100, learning_rate=0.3, max_depth=4, subsample=0.8, n_iter_no_change=2, verbosity=1)\n",
    "xgb_param = xgb_model.get_xgb_params()\n",
    "xgb_param['num_class'] = 7\n",
    "cvresult = xgb.cv(xgb_param, xgb_train, num_boost_round=xgb_model.get_params()['n_estimators'], nfold=5, early_stopping_rounds=10, verbose_eval=True)\n",
    "xgb.set_params(n_estimators=cvresult.shape[0])\n",
    "predictions, accuracy, metrics_report = train_test_model(xgb_model, X_train, X_test, y_train, y_test)\n",
    "print('accuracy: {}'.format(accuracy))\n",
    "print(metrics_report)\n",
    "plot_confusion_matrix(xgb_model, X_test, y_test, display_labels=labels, xticks_rotation='vertical', cmap=\"BuPu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### XGB Implementation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_train = xgb.DMatrix(X_train, label=y_train)\n",
    "xgb_test = xgb.DMatrix(X_test, label=y_test)\n",
    "\n",
    "# setup parameters for xgboost\n",
    "param = {}\n",
    "# use softmax multi-class classification\n",
    "param['objective'] = 'multi:softmax'\n",
    "param['eta'] = 0.05\n",
    "param['max_depth'] = 12\n",
    "param['nthread'] = 4\n",
    "param['num_class'] = 7\n",
    "param['gpu_id'] = 0\n",
    "\n",
    "watchlist = [(xgb_train, 'train'), (xgb_test, 'test')]\n",
    "num_round = 5\n",
    "bst = xgb.train(param, xgb_train, num_round)\n",
    "\n",
    "bst = xgb.train(param, xgb_train, num_round, watchlist)\n",
    "# get prediction\n",
    "pred = bst.predict(xgb_test)\n",
    "error_rate = np.sum(pred != y_test) / y_test.shape[0]\n",
    "print('Test error using softmax = {}'.format(error_rate))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```\n",
    "[0]\ttrain-merror:0.19384\ttest-merror:0.22162\n",
    "[1]\ttrain-merror:0.18778\ttest-merror:0.21685\n",
    "[2]\ttrain-merror:0.18410\ttest-merror:0.21366\n",
    "[3]\ttrain-merror:0.18071\ttest-merror:0.21107\n",
    "[4]\ttrain-merror:0.17842\ttest-merror:0.20849\n",
    "Test error using softmax = 0.208484712984182\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Performance Comparison"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "xx = np.array([1,2,3])\n",
    "width = 0.15\n",
    "gb_prf = np.array([0.83, 0.83, 0.83])\n",
    "ax = plt.subplot(111)\n",
    "ax.bar( xx - width, height=np.array(nb_prf), width=width, color='b', align='center', label='NB', tick_label=['Precision', 'Recall', 'f1'])\n",
    "ax.bar(xx ,height=np.array(log_prf), width=width, color='g', align='center', label='Log_Reg')\n",
    "ax.bar(xx + width, height=np.array(svm_prf), width=width, color='r', align='center', label='SVM')\n",
    "ax.bar(xx + 2*width, height=np.array(rf_prf), width=width, color='y', align='center', label='Random Forest')\n",
    "ax.bar(xx + 3*width, height=np.array(gb_prf), width=width, color='black', align='center', label='Gradient Boosting')\n",
    "\n",
    "\n",
    "plt.xlabel('Classification Metrics')\n",
    "plt.ylabel('Scores')\n",
    "plt.legend(loc='lower right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training Time Comparison\n",
    "\n",
    "The plot below shows that Naive Bayes is incredibly fast and SVM is following it by having the 0.28s and 5.59s training times, respectively. I intentionally omitted Gradient boosting from the plot, because it has very large training time, to be able to have a better visualization."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "tr_times = [('NB', 0.28), ('Log_Reg', 111.02), ('SVM', 5.59), ('Random_Forest', 280.10)]\n",
    "x_vals = []\n",
    "h_vals = []\n",
    "for t in tr_times:\n",
    "    x_vals.append(t[0])\n",
    "    h_vals.append(t[1])\n",
    "    \n",
    "ax = plt.subplot(111)\n",
    "ax.bar(np.linspace(0,1,4) - width, height=np.array(h_vals), width=width, color='Green', tick_label=x_vals)\n",
    "plt.xlabel('Techniques')\n",
    "plt.ylabel('Training Time')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cross Validation Implementation of the Models\n",
    "\n",
    "The second approach to coping with imbalanced datasets is cross validation. For this purpose, I used the *stratified Cross Validation* from [sklearn](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.StratifiedKFold.html#sklearn.model_selection.StratifiedKFold). This method creates the folds by preserving the percentage of samples for each class. Please note that cross validation in general is computationally expensive. Below are the results for all the previous techniques."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Naive Bayes Cross Validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Total time: 1.34s\n",
      "mean accuracy: 0.81\n"
     ]
    }
   ],
   "source": [
    "skfold = StratifiedKFold(n_splits=5, shuffle=True, random_state=123)\n",
    "nb_model = MultinomialNB()\n",
    "\n",
    "accs = []\n",
    "reports = []\n",
    "f1_scores = []\n",
    "st_time = time.time()\n",
    "for train_index, test_index in skfold.split(X,y):\n",
    "    X_train, X_test = X[train_index], X[test_index]\n",
    "    y_train, y_test = y[train_index], y[test_index]\n",
    "    predictions, accuracy, metrics_report, nb_prf = train_test_model(nb_model, X_train, X_test, y_train, y_test, labels)\n",
    "    accs.append(accuracy)\n",
    "    reports.append(metrics_report)\n",
    "    \n",
    "en_time = time.time()\n",
    "print('Total time: {:.2f}s'.format(en_time-st_time))   \n",
    "print('mean accuracy: {:.2f}'.format(np.mean(accs)))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression Cross Validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Total time: 633.33s\n",
      "mean accuracy: 0.85\n"
     ]
    }
   ],
   "source": [
    "log_model = LogisticRegression(penalty='l2', max_iter=500)\n",
    "accs = []\n",
    "reports = []\n",
    "st_time = time.time()\n",
    "for train_index, test_index in skfold.split(X,y):\n",
    "    X_train, X_test = X[train_index], X[test_index]\n",
    "    y_train, y_test = y[train_index], y[test_index]\n",
    "    predictions, accuracy, metrics_report, log_prf = train_test_model(log_model, X_train, X_test, y_train, y_test, labels)\n",
    "    accs.append(accuracy)\n",
    "    reports.append(metrics_report)\n",
    "\n",
    "en_time = time.time()\n",
    "print('Total time: {:.2f}s'.format(en_time-st_time))   \n",
    "print('mean accuracy: {:.2f}'.format(np.mean(accs)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVM Cross Validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Total time: 28.41s\n",
      "mean accuracy: 0.84\n"
     ]
    }
   ],
   "source": [
    "svm = LinearSVC(class_weight='balanced', verbose=False, max_iter=10000, tol=1e-4, C=0.1)\n",
    "accs = []\n",
    "reports = []\n",
    "st_time = time.time()\n",
    "for train_index, test_index in skfold.split(X,y):\n",
    "    X_train, X_test = X[train_index], X[test_index]\n",
    "    y_train, y_test = y[train_index], y[test_index]\n",
    "    predictions, accuracy, metrics_report, svm_prf = train_test_model(svm, X_train, X_test, y_train, y_test, labels)\n",
    "    accs.append(accuracy)\n",
    "    reports.append(metrics_report)\n",
    "\n",
    "en_time = time.time()\n",
    "print('Total time: {:.2f}s'.format(en_time-st_time)) \n",
    "print('mean accuracy: {:.2f}'.format(np.mean(accs)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest Cross Validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Start training...done!\n",
      "Start testing...done!\n",
      "Total time: 1274.23s\n",
      "mean accuracy: 0.84\n"
     ]
    }
   ],
   "source": [
    "rf_model = RandomForestClassifier(n_estimators=100, max_depth=100, min_samples_split=10, n_jobs=-1, verbose=0)\n",
    "accs = []\n",
    "reports = []\n",
    "st_time = time.time()\n",
    "for train_index, test_index in skfold.split(X,y):\n",
    "    X_train, X_test = X[train_index], X[test_index]\n",
    "    y_train, y_test = y[train_index], y[test_index]\n",
    "    predictions, accuracy, metrics_report, rf_prf = train_test_model(rf_model, X_train, X_test, y_train, y_test, labels)\n",
    "    accs.append(accuracy)\n",
    "    reports.append(metrics_report)\n",
    "\n",
    "en_time = time.time()\n",
    "print('Total time: {:.2f}s'.format(en_time-st_time)) \n",
    "print('mean accuracy: {:.2f}'.format(np.mean(accs)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Result Comparison\n",
    "\n",
    "The cross validation results for all the methods are very close. The mean accuracy of random forest and SVM is 84%, logistic regression is still the front runner by 85% accuracy and navie Bayes ranks at the end with 81% accuracy."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Multiclass Classification Using BERT\n",
    "\n",
    "There are several different deep learning techniques available for (text) classification such as sequence to sequence models (LSTM, RNN), Convolutional Neural Networks (CNNs) and their combinations that can produce extremely good results. However, after introducing transformer models in 2017, many transformer-based models developed that led to new state-of-the-art results in a wide variety of NLP tasks including question answering, classification and others. There are many scientific articles demonstrate that such as [this](https://arxiv.org/pdf/1810.04805.pdf) and [this](https://arxiv.org/pdf/1906.08237.pdf). Therefore, I decided to utilize a transformer model for my classification problem. \n",
    "\n",
    "I used [transformers](https://huggingface.co/transformers/index.html) library and built a BERT model. But due to some unknown issue it wasn't training correctly. Therefore, I switched to another library called [ktrain](https://github.com/amaiya/ktrain) and built the model. Please see below for more information."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df['text'].to_list()\n",
    "\n",
    "# Encode the labels\n",
    "labels = df.product_group.unique()\n",
    "label_encoder = encode_labels(labels)\n",
    "y = label_encoder.transform(df.product_group)\n",
    "\n",
    "# Split the dataset\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, stratify=y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## My Implementation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import BertTokenizer, RobertaTokenizer\n",
    "from transformers import TFBertModel, TFBertPreTrainedModel, TFBertForSequenceClassification\n",
    "from transformers import glue_convert_examples_to_features, InputExample\n",
    "\n",
    "def convert_data_into_input_example(x, y):\n",
    "    input_examples = []\n",
    "    for i in tqdm(range(len(x))):\n",
    "        example = InputExample(\n",
    "            guid= None,\n",
    "            text_a= x[i],\n",
    "            text_b= None,\n",
    "            label= str(y[i])\n",
    "        )\n",
    "        input_examples.append(example)\n",
    "    return input_examples\n",
    "\n",
    "def bert_compatiable_format(bdset):  \n",
    "    input_ids, attention_mask, token_type_ids, labels = [], [], [], []\n",
    "    for in_ex in bdset:\n",
    "        input_ids.append(in_ex.input_ids)\n",
    "        attention_mask.append(in_ex.attention_mask)\n",
    "        token_type_ids.append(in_ex.token_type_ids)\n",
    "        labels.append(in_ex.label)\n",
    "\n",
    "    labels = np.vstack(labels)\n",
    "    return ([np.asarray(input_ids), np.asarray(attention_mask), np.asarray(token_type_ids)], labels)\n",
    "\n",
    "tokenizer = BertTokenizer.from_pretrained('bert-base-uncased', do_lower_case=False)\n",
    "train_input_examples = convert_data_into_input_example(X_train, y_train)\n",
    "val_input_examples = convert_data_into_input_example(X_test, y_test)\n",
    "\n",
    "\n",
    "label_list = label_encoder.transform(labels)\n",
    "label_list = [str(i) for i in label_list.tolist()]\n",
    "bert_train_dataset = glue_convert_examples_to_features(examples=train_input_examples, tokenizer=tokenizer, max_length=128, task='mrpc', label_list=label_list)\n",
    "bert_val_dataset = glue_convert_examples_to_features(examples=val_input_examples, tokenizer=tokenizer, max_length=128, task='mrpc', label_list=label_list)\n",
    "\n",
    "x_train, y_train = bert_compatiable_format(bert_train_dataset)\n",
    "x_val, y_val     = bert_compatiable_format(bert_val_dataset)\n",
    "\n",
    "def example_to_features(input_ids, attention_masks, token_type_ids, y):\n",
    "    return {\"input_ids\": input_ids,\n",
    "            \"attention_mask\": attention_masks,\n",
    "            \"token_type_ids\": token_type_ids},y\n",
    "\n",
    "\n",
    "train_ds = tf.data.Dataset.from_tensor_slices((x_train[0], x_train[1], x_train[2], y_train)).map(example_to_features).shuffle(100).batch(64)\n",
    "val_ds   = tf.data.Dataset.from_tensor_slices((x_val[0], x_val[1], x_val[2], y_val)).map(example_to_features).batch(64)\n",
    "model = TFBertForSequenceClassification.from_pretrained('bert-base-uncased')\n",
    "optimizer = tf.keras.optimizers.Adam(learning_rate=3e-5, epsilon=1e-08, clipnorm=1.0)\n",
    "loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "metric = tf.keras.metrics.SparseCategoricalAccuracy('accuracy')\n",
    "model.compile(optimizer=optimizer, loss=loss, metrics=[metric])\n",
    "EPOCHS = 3\n",
    "\n",
    "# Train the model\n",
    "history = model.fit(train_ds, validation_data=val_ds, epochs=EPOCHS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Multiclass Text Classification Using krain\n",
    "\n",
    "I used [ktrain](https://github.com/amaiya/ktrain) library to implement BERT. \"*ktrain is a lightweight wrapper for the deep learning library TensorFlow Keras (and other libraries) to help build, train, and deploy neural networks and other machine learning models. It is designed to make deep learning and AI more accessible and easier to apply for both newcomers and experienced practitioners.*\" "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import ktrain\n",
    "from ktrain import text\n",
    "\n",
    "(x_train,  y_train), (x_test, y_test), preproc = text.texts_from_array(x_train=X_train, y_train=y_train,\n",
    "                                          x_test=X_test, y_test=y_test,\n",
    "                                          class_names=labels.tolist(),\n",
    "                                          preprocess_mode='bert',\n",
    "                                          maxlen=128)\n",
    "model = text.text_classifier('bert', train_data=(x_train, y_train), preproc=preproc)\n",
    "learner = ktrain.get_learner(model, train_data=(x_train, y_train), batch_size=6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# find good learning rate\n",
    "learner.lr_find()             # briefly simulate training to find good learning rate\n",
    "learner.lr_plot()             # visually identify best learning rate\n",
    "learner.fit_onecycle(2e-5, 4)\n",
    "\n",
    "# Use a Predictor object capable of making predictions on new raw data.\n",
    "predictor = ktrain.get_predictor(learner.model, preproc)\n",
    "\n",
    "predictor.get_classes()\n",
    "\n",
    "# Predict a new document\n",
    "predictor.predict(X_test[0:1])\n",
    "\n",
    "# let's save the predictor for later use\n",
    "predictor.save('/tmp/my_predictor')\n",
    "\n",
    "# reload the predictor\n",
    "reloaded_predictor = ktrain.load_predictor('/tmp/my_predictor')\n",
    "\n",
    "# make a prediction on the same document to verify it still works\n",
    "reloaded_predictor.predict(x_test[0:1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CPU Run\n",
    "\n",
    "```\n",
    "begin training using onecycle policy with max lr of 2e-05...\n",
    "Train on 214688 samples\n",
    "Epoch 1/4\n",
    "  4530/214688 [..............................] - ETA: 18:36:08 - loss: 1.4979 - accuracy: 0.4507\n",
    "  ```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## GPU Run\n",
    "\n",
    "I used the Google colab to utilize the GPU resource. The program started with no problem. However, at the end of the second epoch due to lack of GPU resource it stopped and restarted the environment. Every epoch took 1.30 hours to execute and it can be observed that the accuracy is quite good (87%). I think it would go beyond 90% if it wasn't stopped. \n",
    "\n",
    "```\n",
    "\n",
    "begin training using onecycle policy with max lr of 3e-05...\n",
    "Train on 214688 samples\n",
    "Epoch 1/3\n",
    "214688/214688 [==============================] - 10178s 47ms/sample - loss: 0.5175 - accuracy: 0.8252\n",
    "Epoch 2/3\n",
    "214688/214688 [==============================] - 10165s 47ms/sample - loss: 0.3903 - accuracy: 0.8684\n",
    "Epoch 3/3\n",
    "214688/214688 [==============================] - 10200s 48ms/sample - loss: 0.2637 - accuracy: 0.9110\n",
    "<tensorflow.python.keras.callbacks.History at 0x7f59883c4668>\n",
    "\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Challenges and Discussion\n",
    "\n",
    "This project had several challenges including:\n",
    "\n",
    "- The dataset was fairly large, which made it quite interesting.\n",
    "\n",
    "- The dataset was imbalanced in terms of number of documents in different classes. Also, the length of documents varied from 1 to over 5000 words.\n",
    "\n",
    "- The textual content needed plenty of cleaning\n",
    "\n",
    "\n",
    "Since I did not have easy access to GPU resources, I wasn't able to get the result that I expected. That said, The best result among all the six models I trained belongs to BERT by **91%** accuracy.\n",
    "\n",
    "In conclusion, transformer BERT outperformed all other models (even though it wasn't completely finished!!!), but training time was significantly higher and needs GPU. Naive Bayes on the other hand, is exceptionally fast and gives reasonable results. SVM and logistic regression were in the middle, fast in training and high accuracy, respectively. Random forest and GBoosting methods were the slowest with less accuracy. Standard/traditonal methods need feature engineering to improve perforamnce, but transformer models are pretrained and can be used end to end."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}