01-Eda-baselines.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h1>Machine Learning to Identify Toxic Comments: EDA & Baselines</h1>\n",
    "\n",
    "### Breif Overview of Problem Statement:\n",
    "\n",
    "With too much freedom in social media, profanity in comments (i.e. comments that are disrespectful, rude or otherwise likely to make someone leave a discussion) can be a serious issue if not addressed. Moreover, with surge in number users accessing social media these days, this task cannot be done manually with help of moderators. So to automate the identification of toxic comments on social media platforms, The Conversation AI team had come up with an interesting kaggle competition to build a roboust machine learning model that can do this task.\n",
    "\n",
    "For this competition, we are provided a annotated dataset of comments from Wikipedia’s talk page edits. Using this as training data, we are challenged to build a multi-headed machine learning model that is capable of detecting different types of of toxicity like obscenity, threats, insults, and identity-based hate better than existing machine learning models (yes, conversational AI team already has a model to do this).\n",
    "\n",
    "More details of the competition can be found here: \n",
    "> source: [kaggle](https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge)\n",
    "\n",
    "\n",
    "### Data & Evaluation metrics:\n",
    "\n",
    "`train.csv` contains a large number of Wikipedia comments (around 100K+) which have been labeled by human raters for toxic behavior. The types of toxicity are:\n",
    "\n",
    "`toxic\n",
    " severe_toxic\n",
    " obscene\n",
    " threat\n",
    " insult\n",
    " identity_hate`\n",
    " \n",
    "- Apart from the above mentioned classes there will be also completely clean comments.\n",
    "- The trained model will be tested on `test.csv` data using mean column-wise ROC AUC as evaluation metric.\n",
    "- **Mean column-wise ROC AUC = Average of the individual ROC AUC scores of each predicted class. **\n",
    "\n",
    "Submission File Format:\n",
    "\n",
    "For each id in the test set, you must predict a probability for each of the six possible types of comment toxicity (toxic, severetoxic, obscene, threat, insult, identityhate). The columns must be in the same order as shown below. The file should contain a header and have the following format:\n",
    "\n",
    "`id,toxic,severe_toxic,obscene,threat,insult,identity_hate\n",
    " 00001cee341fdb12,0.5,0.5,0.5,0.5,0.5,0.5\n",
    " 0000247867823ef7,0.5,0.5,0.5,0.5,0.5,0.5\n",
    " etc.`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Objective of this Notebook: \n",
    "- Explore the data, perform analysis & build baselines.\n",
    "\n",
    "### Reading the Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train data shape: (159571, 8)\n",
      "Columns in Train: Index(['id', 'comment_text', 'toxic', 'severe_toxic', 'obscene', 'threat',\n",
      "       'insult', 'identity_hate'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import os\n",
    "data_paths = {}\n",
    "for dirname, _, filenames in os.walk('/kaggle/input'):\n",
    "    for filename in filenames:\n",
    "        data_paths[filename] = os.path.join(dirname, filename)\n",
    "        # print(os.path.join(dirname, filename))\n",
    "        \n",
    "train_df = pd.read_csv(data_paths['train.csv'])\n",
    "test_df = pd.read_csv(data_paths['test.csv'])\n",
    "sub_df = pd.read_csv(data_paths['sample_submission.csv'])\n",
    "print('Train data shape:', train_df.shape)\n",
    "print('Columns in Train:', train_df.columns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The provided `train.csv` is having 159571 rows and 8 columns. Let us now sample to take 5 rows from the data to get visual understanding of what is present."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>24915</th>\n",
       "      <td>41e65f43c180838e</td>\n",
       "      <td>YOU ARE A FAT, GEEKY PRICK WHO HAS NOTHING TO ...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75819</th>\n",
       "      <td>cade9e425d75948e</td>\n",
       "      <td>Agent X2: Basically thanks - with a 'little' m...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53891</th>\n",
       "      <td>8ff87747403ed2e6</td>\n",
       "      <td>Why are my posts being deleted? \\n\\nI have tri...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>154159</th>\n",
       "      <td>a95604c7a938abb3</td>\n",
       "      <td>\"\\n\\n Controlled Demolitions and Common Sense ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13040</th>\n",
       "      <td>2288910537399728</td>\n",
       "      <td>I do not understand your reply.  //Blaxthos ( ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                      id                                       comment_text  \\\n",
       "24915   41e65f43c180838e  YOU ARE A FAT, GEEKY PRICK WHO HAS NOTHING TO ...   \n",
       "75819   cade9e425d75948e  Agent X2: Basically thanks - with a 'little' m...   \n",
       "53891   8ff87747403ed2e6  Why are my posts being deleted? \\n\\nI have tri...   \n",
       "154159  a95604c7a938abb3  \"\\n\\n Controlled Demolitions and Common Sense ...   \n",
       "13040   2288910537399728  I do not understand your reply.  //Blaxthos ( ...   \n",
       "\n",
       "        toxic  severe_toxic  obscene  threat  insult  identity_hate  \n",
       "24915       1             1        1       0       1              0  \n",
       "75819       0             0        0       0       0              0  \n",
       "53891       0             0        0       0       0              0  \n",
       "154159      0             0        0       0       0              0  \n",
       "13040       0             0        0       0       0              0  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.sample(5, random_state = 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- This is going to be text classification problem...we are given a comment and we need to classify whether it is toxic.\n",
    "- The column `comment_text` contains the comments text and it will be our main feature.\n",
    "- All the remaining columns except `id` will be our target variable to predict.\n",
    "- Target varaibles are binary encoded (one hot encoded) & all zeros represents a clean comment.\n",
    "- A single comments can belong to multiple toxicity classes\n",
    "\n",
    "Now lets look into few complete random comments from the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\"\\nWell, not \"\"before the process\"\" but \"\"before how we do things with subpages\"\" His RfA is listed on NoSeptember\\'s page and you can find it if you look. September 2004 I think. I have my differences with El_C to be sure, but was surprised to see a block, so I left a note. ++: t/c \"'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'They are deletionists.'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "\"Hello \\n\\nHello, I'm Irene.\\n\\nI created See Me novel from Nicholas Sparks. I created it because I saw Nicholas Spark's official website about a new novel released coming soon. I ask you, is this page wrong or anything?\\n\\nThanks.\""
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in [34, 55345, 124786]:\n",
    "    display(train_df.loc[i, 'comment_text'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Clearly, comments are largely varying in lengths. Let us see how their distrubution looks like"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Central Tendencies on lengths of comment_text\n",
      " count    159571.000000\n",
      "mean        394.073221\n",
      "std         590.720282\n",
      "min           6.000000\n",
      "25%          96.000000\n",
      "50%         205.000000\n",
      "75%         435.000000\n",
      "max        5000.000000\n",
      "Name: comment_text, dtype: float64\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "comment_lens = train_df['comment_text'].str.len()\n",
    "print('Central Tendencies on lengths of comment_text\\n', comment_lens.describe())\n",
    "ax = comment_lens.hist()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From the above outputs we can say these points:\n",
    "\n",
    "- Comments have no null values.\n",
    "- Mininum length of a comment is only 6 chars while maximum length can be 5000 characters.\n",
    "- Mean (394 chars) and median (205 chars) are not close indicating the skewness - as a result histogram is left sided\n",
    "\n",
    "Now we will look how many of given comments are clean & how many are toxic (in other words, distrubution of classes to be predicted)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Clean comments (All 0's in a target row) in train: 143346\n",
      "Total unclean/bad comments (atleast one 1 in a target row)in train: 16225\n",
      "Total label tags (total counts of 1's in target columns): 35098\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "drop_col = ['id', 'is_clean']  # columns not neccessary - can be dropped\n",
    "text_col = ['comment_text']  # text feature\n",
    "label_col = [col for col in train_df.columns if col not in text_col + drop_col] # target variables\n",
    "\n",
    "labels_per_comment = train_df[label_col].sum(axis = 1) # clac no.of labels for each comment\n",
    "\n",
    "# add a new column to indicate if a comment is toxic (bad) or not (clean).\n",
    "train_df['is_clean'] = 0\n",
    "train_df.loc[labels_per_comment == 0, 'is_clean'] = 1\n",
    "# train_df['is_clean'].value_counts()\n",
    "\n",
    "print(\"Total Clean comments (All 0's in a target row) in train:\",len(train_df[train_df['is_clean'] == 1]))\n",
    "print(\"Total unclean/bad comments (atleast one 1 in a target row)in train:\",len(train_df[train_df['is_clean'] != 1]))\n",
    "print(\"Total label tags (total counts of 1's in target columns):\",train_df[label_col].sum().sum())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Importantly we need to notice that total bad comments & total label tags are not equal, so said intially there are multiple labels for many comments. On Average, there are 2 labels for each comment - but let us see - Now let us see how multiple labels are divided:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "tags_count = labels_per_comment.value_counts()\n",
    "\n",
    "# plotting the label counts\n",
    "plt.figure(figsize=(8,4))\n",
    "ax = sns.barplot(tags_count.index, tags_count.values, alpha=0.7)\n",
    "plt.title(\"Tags Counts v/s Occurences in Train Data\")\n",
    "plt.ylabel('# of Occurrences', fontsize=12)\n",
    "plt.xlabel('# Tag Count', fontsize=12)\n",
    "\n",
    "#adding the text labels\n",
    "rects = ax.patches\n",
    "labels = tags_count.values\n",
    "for rect, label in zip(rects, labels):\n",
    "    height = rect.get_height()\n",
    "    ax.text(rect.get_x() + rect.get_width()/2, height + 10, \n",
    "            label, ha='center', va='bottom')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- maximum no.of tags for a bad comment is 6 (i.e it will belong to all six classes).\n",
    "- This we can be viewed as highericial tagging. (Ex: A comment which is `severe_toxic` will be also a `toxic`).\n",
    "\n",
    "Lets visualize how the main six classes are distributed (counts of classes) among the 35098 occurences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "label_counts = train_df[label_col].sum()\n",
    "\n",
    "# plotting the label counts\n",
    "plt.figure(figsize=(8,4))\n",
    "ax = sns.barplot(label_counts.index, label_counts.values, alpha=0.7)\n",
    "plt.title(\"Counts Per Class\")\n",
    "plt.ylabel('# of Occurrences', fontsize=12)\n",
    "plt.xlabel('Label', fontsize=12)\n",
    "\n",
    "#adding the text labels\n",
    "rects = ax.patches\n",
    "labels = label_counts.values\n",
    "for rect, label in zip(rects, labels):\n",
    "    height = rect.get_height()\n",
    "    ax.text(rect.get_x() + rect.get_width()/2, height + 10, \n",
    "            label, ha='center', va='bottom')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The above figure tells us that Toxicity type is not evenly spreadout - `toxic` comments are more common while `threat` is least occuring class. This is interesting & a bit scary too. In total, there is a high imbalance b/w the clean & unclean comments. Again with in unclean comments, there is imbalance b/w the types of toxicity!\n",
    "\n",
    "Till now we only analyzed the target varaibles, now we will look at one comment each from the all 6 types of toxic classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "toxic - comment sample :\n",
      "Rjanag, your foul language showed your true white trash mentality. You being white prove more of your bias and hideous intent. Keep your effort in trying to hide the truth. If you can certainly call up your gang to block people from Wikipedia, it only serves as proof of your fear of the truth.\n",
      "\n",
      "--------------------------------------------------\n",
      "\n",
      "severe_toxic - comment sample :\n",
      "what i mean to say is that im a big fucknig idiot and i dont know what im talking about\n",
      "\n",
      "--------------------------------------------------\n",
      "\n",
      "obscene - comment sample :\n",
      "Warning\n",
      "Get off your high horse with warnings and referring to your talk page as if it were another user. You're no admin, but like the other assclowns that have gotten on my case, you seem to think you've got cred by throwing warnings at me. No one is gonna check the talk page of a low importance-rated article and you know it, you dumbfuck.\n",
      "\n",
      "--------------------------------------------------\n",
      "\n",
      "threat - comment sample :\n",
      "I'll kick the shit out of you ya cunt\n",
      "\n",
      "--------------------------------------------------\n",
      "\n",
      "insult - comment sample :\n",
      "Speculate? A retarded 10 year old child can see what is happening. 122.60.93.162\n",
      "\n",
      "--------------------------------------------------\n",
      "\n",
      "identity_hate - comment sample :\n",
      "Hi, you poor twerps. My money does far more good than your pathetic rule enforcing. Have fun with your fat wives and gay hookers.\n",
      "\n",
      "--------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "import random\n",
    "for label in label_col:\n",
    "    label_df = train_df[train_df[label]==1].reset_index(drop = 1)\n",
    "    print('\\n' + label + ' - comment sample :')\n",
    "    print(label_df.loc[random.randint(0, len(label_df)-1), 'comment_text'])\n",
    "    print('\\n' + '-'*50)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Aah...that's a lot of profanity...!\n",
    "- Comments are having puntuctions & things user want to highlight in captials (or) quotes.\n",
    "- As usually, Contractions are always there. In some cases (very less), there is also text from other languages in b/w english.\n",
    "- comments include lots of chatting/social shortcuts which we generally use on social media.\n",
    "- few comments seems to contain numbers, time and also IP address (might be of user's).\n",
    "- In few cases there are also Device Id's and Urls posted by users.\n",
    "- Some comments are spam - repeating same things 10's of times.\n",
    "- In few rare cases, I found the word `unblock` before the comment - this might indicate that comment is blocked as it is toxic."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, let us start exploring into comment texts with some questions which relates to our objective. Our objective here is to classify the bad comments into different to toxicity types.\n",
    "\n",
    "- Are Longer comments more toxic ?\n",
    "- did Presence of special characters vary with Toxicity ?\n",
    "- Are spammer more toxic ?\n",
    "\n",
    "\n",
    "These questions will answer how Toxicity of comments vary with different features of comment_text like length, presence of punctuations, capital letters, words, or sentences."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x432 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import re\n",
    "\n",
    "#Total chars:\n",
    "train_df['total_len'] = train_df['comment_text'].apply(len)\n",
    "test_df['total_len'] = test_df['comment_text'].apply(len)\n",
    "\n",
    "#Sentence count in comment: '\\n' is split & count number of sentences in each comment\n",
    "train_df['sent_count'] = train_df[\"comment_text\"].apply(lambda x: len(re.findall(\"\\n\",str(x)))+1)\n",
    "test_df['sent_count'] = test_df[\"comment_text\"].apply(lambda x: len(re.findall(\"\\n\",str(x)))+1)\n",
    "\n",
    "#Word count in each comment:\n",
    "train_df['word_count'] = train_df[\"comment_text\"].apply(lambda x: len(str(x).split()))\n",
    "test_df['word_count'] = test_df[\"comment_text\"].apply(lambda x: len(str(x).split()))\n",
    "\n",
    "\n",
    "plt.figure(figsize=(18,6))\n",
    "plt.suptitle(\"Are longer comments more toxic?\",fontsize=18)\n",
    "plt.tight_layout()\n",
    "\n",
    "# total lengths (characters)\n",
    "plt.subplot(131)\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 0].total_len, label=\"UnClean\",shade=True,color='r')\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 1].total_len, label=\"Clean\")\n",
    "plt.legend()\n",
    "plt.ylabel('Number of occurances', fontsize=12)\n",
    "plt.xlabel('# of Chars', fontsize=12)\n",
    "# plt.title(\"# Chars v/s Toxicity\", fontsize=12)\n",
    "\n",
    "# words\n",
    "plt.subplot(132)\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 0].word_count, label=\"UnClean\",shade=True,color='r')\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 1].word_count, label=\"Clean\")\n",
    "plt.legend()\n",
    "plt.xlabel('# of Words', fontsize=12)\n",
    "# plt.title(\"# Words v/s comment Toxicity\", fontsize=12)\n",
    "\n",
    "## sentences\n",
    "plt.subplot(133)\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 0].sent_count, label=\"UnClean\",shade=True,color='r')\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 1].sent_count, label=\"Clean\")\n",
    "plt.legend()\n",
    "plt.xlabel('# of Sentences', fontsize=12)\n",
    "# plt.title(\"# Sentences v/s comment Toxicity\", fontsize=12)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- most comments are having less than 25 sentences & less than 250 words\n",
    "- unclean comments are having more no.of words in less no.of sentences.\n",
    "- The distrubution plots of clean & unclean of all three plots are very much overlapping with each others, indicating these features are going to be less significant in differentiating them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x432 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import string\n",
    "\n",
    "#Captial letters:\n",
    "train_df['capitals'] = train_df['comment_text'].apply(lambda x: sum(1 for c in x if c.isupper()))\n",
    "test_df['capitals'] = test_df['comment_text'].apply(lambda x: sum(1 for c in x if c.isupper()))\n",
    "\n",
    "# #Captials ratio:\n",
    "# train_df['capitals_percent'] = train_df['capitals']/train_df['total_len'] * 100\n",
    "# test_df['capitals_percent'] = test_df['capitals']/train_df['total_len'] * 100\n",
    "\n",
    "# punct count:\n",
    "train_df['punct_count'] = train_df['comment_text'].apply(lambda x: sum(1 for c in x if c in string.punctuation))\n",
    "test_df['punct_count'] = test_df['comment_text'].apply(lambda x: sum(1 for c in x if c in string.punctuation))\n",
    "\n",
    "# smilies:\n",
    "smilies = (':-)', ':)', ';-)', ';)')\n",
    "train_df['smilies_count'] = train_df['comment_text'].apply(lambda comment: sum(comment.count(s) for s in smilies))\n",
    "test_df['smilies_count'] = test_df['comment_text'].apply(lambda comment: sum(comment.count(s) for s in smilies))\n",
    "\n",
    "#----------plotting------------\n",
    "\n",
    "plt.figure(figsize=(18,6))\n",
    "plt.suptitle(\"did Presence of special characters vary with Toxicity ?\\n\",fontsize=18)\n",
    "plt.tight_layout()\n",
    "\n",
    "# words\n",
    "plt.subplot(131)\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 0].capitals, label=\"UnClean\",shade=True,color='r')\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 1].capitals, label=\"Clean\")\n",
    "plt.legend()\n",
    "plt.ylabel('Number of occurances', fontsize=12)\n",
    "plt.xlabel('# Capital letters', fontsize=12)\n",
    "# plt.title(\"# Captials v/s Toxicity\", fontsize=12)\n",
    "\n",
    "# words\n",
    "plt.subplot(132)\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 0].punct_count, label=\"UnClean\",shade=True,color='r')\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 1].punct_count, label=\"Clean\")\n",
    "plt.legend()\n",
    "plt.xlabel('# of Punctuations', fontsize=12)\n",
    "# plt.title(\"#Punctuations v/s comment Toxicity\", fontsize=12)\n",
    "\n",
    "## sentences\n",
    "plt.subplot(133)\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 0].smilies_count, label=\"UnClean\",shade=True,color='r')\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 1].smilies_count, label=\"Clean\")\n",
    "plt.legend()\n",
    "plt.xlabel('# of Smilies', fontsize=12)\n",
    "# plt.title(\"#Smilies v/s comment Toxicity\", fontsize=12)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- presence of captial letters is more in case of unclean comments, but the distrbutions are overlapping making it a difficult feature for models to extract information.\n",
    "- most of the clean comments are having punctuations less than 100 while for unclean comments it spread to max of 5000 punctuations.\n",
    "- no.of smilies in unclean v/s clean comments is very much similar and unclean comments are having more comments with no.of smilies = 1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Unique word count:\n",
    "train_df['unique_word_count'] = train_df[\"comment_text\"].apply(lambda x: len(set(str(x).split())))\n",
    "test_df['unique_word_count'] = test_df[\"comment_text\"].apply(lambda x: len(set(str(x).split())))\n",
    "\n",
    "#Unique ratio:\n",
    "train_df['unique_word_percent'] = train_df['unique_word_count']/train_df['word_count'] * 100\n",
    "test_df['unique_word_percent'] = test_df['unique_word_count']/train_df['word_count'] * 100\n",
    "\n",
    "#----------plotting------------\n",
    "\n",
    "# comments with unique word count percentage < 25%...they can be spam/referal links/marketing links\n",
    "\n",
    "plt.figure(figsize=(15,5))\n",
    "plt.suptitle(\"Comments with less-unique-words(spam) are more toxic?\",fontsize = 18)\n",
    "\n",
    "plt.subplot(121)\n",
    "plt.title(\"% of unique words in comments\")\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 0].unique_word_percent, label=\"UnClean\",shade=True,color='r')\n",
    "ax=sns.kdeplot(train_df[train_df.is_clean == 1].unique_word_percent, label=\"Clean\")\n",
    "plt.legend()\n",
    "plt.ylabel('Number of occurances', fontsize=12)\n",
    "plt.xlabel('Percent unique words', fontsize=12)\n",
    "\n",
    "plt.subplot(122)\n",
    "sns.violinplot(y = 'unique_word_count',x='is_clean', data = train_df[train_df['unique_word_percent'] < 25], \n",
    "               split=True,inner=\"quart\")\n",
    "plt.xlabel('is_Clean', fontsize=12)\n",
    "plt.ylabel('# of words', fontsize=12)\n",
    "plt.title(\"# unique words v/s Toxicity\")\n",
    "plt.show()\n",
    "\n",
    "# train_df[train_df['word_unique_percent'] < 25]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- There is a wide spread area for unclean points in the unique word percentage range of 1-10%, Interesting there are clean comments as well with lesser number of unique words. \n",
    "- This feature seems carry some significance especially incase of sentences with less unique words.\n",
    "- lets once see how text in clean-spam & unclean-spam comments look like"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Clean Spam example:\n",
      "MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD MUD\n",
      "--------------------------------------------------\n",
      "Toxic Spam example:\n",
      "I OWN THIS PAGE YOU FOOLS!! YOU SHOULD FEAR ME!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEORGE BUSH SHOULD GO TO HELL!!!!!!!!! I AM THE GOD OF CHOCOBOS, I WANT MY CHEESE, GEOR\n"
     ]
    }
   ],
   "source": [
    "## lets have a look how clean & unclean spam comment looks like\n",
    "\n",
    "print(\"Clean Spam example:\")\n",
    "print(train_df[train_df['unique_word_percent'] < 10][train_df['is_clean'] == 1].comment_text.iloc[3])\n",
    "print('-'*50)\n",
    "print(\"Toxic Spam example:\")\n",
    "print(train_df[train_df['unique_word_percent'] < 10][train_df['is_clean'] == 0].comment_text.iloc[25])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- we definitely need to penalize the over repeating words (or) there will be chance for models to learn as words like 'George Bush' as indicator as toxicity. so comments with lesser unique words should be preprocessed carefully.\n",
    "\n",
    "- Tfidf is well known for this penalizing effect.\n",
    "-  TF-IDF Vectorizer\n",
    "      - TF - Term Frequency -- Count of the words(Terms) in the text corpus (same of Count Vect)\n",
    "      - IDF - Inverse Document Frequency -- Penalizes words that are too frequent. We can think of this as regularization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "_kg_hide-input": true
   },
   "outputs": [],
   "source": [
    "train_df.to_csv('train_feateng.csv', index = None)\n",
    "test_df.to_csv('test_feateng.csv', index = None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "_kg_hide-input": true
   },
   "outputs": [],
   "source": [
    "# from nltk.corpus import stopwords\n",
    "# from nltk import pos_tag\n",
    "# from nltk.stem.wordnet import WordNetLemmatizer \n",
    "# from nltk.tokenize import word_tokenize # Tweet tokenizer does not split at apostophes which is what we want\n",
    "# from nltk.tokenize import TweetTokenizer\n",
    "\n",
    "\n",
    "# lemma = WordNetLemmatizer()\n",
    "# tokenizer=TweetTokenizer()\n",
    "# eng_stopwords = list(stopwords.words('english'))\n",
    "\n",
    "# def simple_preprocess(comment):\n",
    "#     \"\"\"\n",
    "#     This function receives comments and returns clean word-list\n",
    "#     \"\"\"\n",
    "#     #Convert to lower case \n",
    "#     comment=comment.lower()\n",
    "#     #remove \\n\n",
    "#     comment=re.sub(\"\\\\n\",\"\",comment)\n",
    "\n",
    "#     #Split the sentences into words and lemmatize\n",
    "#     words=tokenizer.tokenize(comment)\n",
    "#     words=[lemma.lemmatize(word, \"v\") for word in words]\n",
    "#     words = [w for w in words if not w in eng_stopwords]\n",
    "    \n",
    "#     clean_sent=\" \".join(words)\n",
    "    \n",
    "#     return(clean_sent)\n",
    "\n",
    "# train_df['comment_text'] = train_df['comment_text'].apply(lambda x: simple_preprocess(x))\n",
    "# test_df['comment_text'] = test_df['comment_text'].apply(lambda x: simple_preprocess(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_topn_tfidf_feat_byClass(X_tfidf, y_train, feature_names, labels, topn):\n",
    "    \n",
    "    feat_imp_dfs = {}\n",
    "    \n",
    "    for label in labels:\n",
    "        # get indices of rows where label is true\n",
    "        label_ids = y_train.index[y_train[label] == 1]\n",
    "        # get subset of rows\n",
    "        label_rows = X_tfidf[label_ids].toarray()\n",
    "        # calc mean feature importance\n",
    "        feat_imp = label_rows.mean(axis = 0)\n",
    "        # sort by column dimension and get topn feature indices\n",
    "        topn_ids = np.argsort(feat_imp)[::-1][:topn]\n",
    "        # combine tfidf value with feature name\n",
    "        topn_features = [(feature_names[i], feat_imp[i]) for i in topn_ids]\n",
    "        # df\n",
    "        topn_df = pd.DataFrame(topn_features, columns = ['word_feature', 'tfidf_value'])\n",
    "        # save \n",
    "        feat_imp_dfs[label] = topn_df\n",
    "    return feat_imp_dfs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Transforming Text Data into TD-IDF vectors & Analysing TF-IDF values \n",
    "\n",
    "#### 1. Visualising Bigram features per Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from nltk.corpus import stopwords\n",
    "from nltk import pos_tag\n",
    "from nltk.stem.wordnet import WordNetLemmatizer \n",
    "from nltk.tokenize import word_tokenize # Tweet tokenizer does not split at apostophes which is what we want\n",
    "from nltk.tokenize import TweetTokenizer\n",
    "\n",
    "\n",
    "tfidf = TfidfVectorizer(ngram_range = (1,1), min_df = 100, \n",
    "                        strip_accents='unicode', analyzer='word',\n",
    "                        use_idf=1,smooth_idf=1,sublinear_tf=1,\n",
    "                        stop_words = 'english')\n",
    "X_unigrams = tfidf.fit_transform(train_df['comment_text'])\n",
    "X_unigrams.shape, len(tfidf.get_feature_names())\n",
    "\n",
    "\n",
    "feature_names = np.array(tfidf.get_feature_names())\n",
    "imp_dfs = get_topn_tfidf_feat_byClass(X_unigrams, train_df, feature_names, label_col, topn = 10)\n",
    "\n",
    "plt.figure(figsize=(15,10))\n",
    "\n",
    "for i, label in enumerate(label_col):\n",
    "    plt.subplot(3, 2, i + 1)\n",
    "    sns.barplot(imp_dfs[label].word_feature[:10], imp_dfs[label].tfidf_value[:10], alpha = 0.8)\n",
    "    plt.title(\"Important UniGrams for the class:{}\".format(label))\n",
    "    plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- few words like `fuck` seems to be in every class, but again as this is multi-label classification (multiple tags for each comment) there will be that overlapping.\n",
    "- especially `threat` class is standing apart with words like `kill`, `die`, `death`.\n",
    "- Interestingly, due to high tf-idf value, word `wikipedia` has stands in top10 features for the `toxic` class. which model should not learn."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2. Visualising Bigram features per Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x864 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "tfidf = TfidfVectorizer(ngram_range = (2,2), min_df = 100, \n",
    "                        strip_accents='unicode', analyzer='word',\n",
    "                        use_idf=1,smooth_idf=1,sublinear_tf=1,\n",
    "                        stop_words = 'english')\n",
    "X_bigrams = tfidf.fit_transform(train_df['comment_text'])\n",
    "X_bigrams.shape, len(tfidf.get_feature_names())\n",
    "\n",
    "feature_names = np.array(tfidf.get_feature_names())\n",
    "imp_dfs = get_topn_tfidf_feat_byClass(X_bigrams, train_df, feature_names, label_col, topn = 10)\n",
    "\n",
    "plt.figure(figsize=(15,12))\n",
    "\n",
    "for i, label in enumerate(label_col):\n",
    "    plt.subplot(3, 2, i + 1)\n",
    "    by_class = sns.barplot(imp_dfs[label].word_feature[:10], imp_dfs[label].tfidf_value[:10], alpha = 0.8)\n",
    "    plt.title(\"Important BiGrams for the class:{}\".format(label))\n",
    "    for item in by_class.get_xticklabels():\n",
    "        item.set_rotation(45)\n",
    "    plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- the top 10 tfidf features in case are not much differentiating, every class are having almost same phrases.\n",
    "- remember, the `comment_text` is completely unprocessed for now.\n",
    "- There is high chance for Names of persons like famous names `george bush`, `mitt romney` (if mentioned repeatedly in the spam) to appear in the bigram features for any class. To avoid this we can preprocess spam comments (comments with < 25% unique words) differently like removing usernames/person names from the comments using regular expressions or POS tags."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Splitting the data into Training & Validation sets:\n",
    "\n",
    "Before testing the model on the actual test data, we neeed to cross-validate internally that model is trained in a right way. Validation set will also help to tune model properly. I am using sklearn's `train_test_split` utility function to do this random splitting. I am using `random_state = 2019`as seed for purpose of reproducability & `test_size` = 0.2 meaning 20% of data points in the train set will be used as validation data and remaining 80% of data will be used to train the model.\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data points in train data after splitting: 127656\n",
      "Data points in valiadtion data: 31915\n",
      "Data points in test data: 153164\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_val, y_train, y_val = train_test_split(train_df['comment_text'], \n",
    "                                                  train_df[label_col], test_size=0.2, random_state=2019)\n",
    "X_test = test_df['comment_text']\n",
    "print('Data points in train data after splitting:', len(X_train))\n",
    "print('Data points in valiadtion data:', len(X_val))\n",
    "print('Data points in test data:', len(X_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Naive Baselines:\n",
    "\n",
    "Before getting to actual modelling & algorithms. let me build some naive baseline on validation set to get minimal optimal score. These scores will be useful for analysing & selection of the models we will be building.\n",
    "\n",
    "#### naive baseline 1: Random guessing probabilities b/w 0 & 1.\n",
    "\n",
    "This time instead having common probility...we take random value for probability for every column in row. What good is a machine learing if its results are not even better than a random prediction ?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Naive Baseline: Random Guessing\n",
      "ROC-AUC score : 0.5075030929045684\n",
      "Log Loss: 0.4697888746965487\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import log_loss, roc_auc_score\n",
    "y_val_naive1 = np.random.rand(y_val.shape[0], y_val.shape[1])\n",
    "print('Naive Baseline:', 'Random Guessing')\n",
    "print('ROC-AUC score :', roc_auc_score(y_val, y_val_naive1))\n",
    "print('Log Loss:', log_loss(y_val, y_val_naive1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Naive baseline 2: Predicting all probabilities as 0.5\n",
    "- Here we are guessing more intelligently...as every comments in validation set as clean. From data analysi we found that train data is highly imbalanced there is high chance that validation set will be also have same distrbutions and predicting every comment as clean will actually give us 90% accuracy...this is also reason why accuracy is not an evalution metric for this problem."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Naive Baseline: Random Guessing\n",
      "ROC-AUC score : 0.5\n",
      "Log Loss: 0.40652139485133465\n"
     ]
    }
   ],
   "source": [
    "y_val_naive2 = np.zeros(y_val.shape)\n",
    "y_val_naive2[:] = 0.5\n",
    "print('Naive Baseline:', 'Random Guessing')\n",
    "print('ROC-AUC score :', roc_auc_score(y_val, y_val_naive2))\n",
    "print('Log Loss:', log_loss(y_val, y_val_naive2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- So the machine model we are going to build should atleast have validation log loss < 0.40 & roc-auc > 0.5. which seems not that tough to achieve.\n",
    "- Before starting the modelling,I am transforming the train, valid & test data splits into tfidf features with both unigrams & bigrams."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final Data dimensions after transformations: (127656, 52083) (127656, 6) (31915, 52083) (31915, 6)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_val, y_train, y_val = train_test_split(train_df['comment_text'], \n",
    "                                                  train_df[label_col], test_size=0.2, random_state=2019)\n",
    "X_test = test_df['comment_text']\n",
    "\n",
    "tfidf = TfidfVectorizer(ngram_range = (1,2), min_df = 9, strip_accents='unicode', analyzer='word',\n",
    "                        use_idf=1, smooth_idf=1, sublinear_tf=1,stop_words = 'english')\n",
    "X_train_tf = tfidf.fit_transform(X_train)\n",
    "X_val_tf = tfidf.transform(X_val)\n",
    "X_test_tf = tfidf.transform(X_test)\n",
    "feature_names = tfidf.get_feature_names()\n",
    "\n",
    "print('Final Data dimensions after transformations:', X_train_tf.shape, y_train.shape, X_val_tf.shape, y_val.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Modelling: Baselines\n",
    "\n",
    "Generally, I will begin with least complex models & will move to complex models based on their time/space complexity.\n",
    "\n",
    "#### 1. Naive Bayes\n",
    "- Naive Bayes is well know for its Text Sentiment Classification tasks. It is also simple to interpret the model so I am starting with it.\n",
    "- If works well, added advantages includes speed - as worst case complexity is O(Nd) is comparatively low.\n",
    "      - N = No.of data points\n",
    "      - d = dimensions (features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Naive Bayes\n",
      "mean ROC-AUC on train set: 0.9447035800357547\n",
      "mean ROC-AUC on validation set: 0.9047281844576962\n"
     ]
    }
   ],
   "source": [
    "from sklearn.multiclass import OneVsRestClassifier\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "model = OneVsRestClassifier(MultinomialNB(), n_jobs = -1)\n",
    "model.fit(X_train_tf, y_train)\n",
    "print('model: Naive Bayes')\n",
    "print('mean ROC-AUC on train set:', roc_auc_score(y_train, model.predict_proba(X_train_tf)))\n",
    "y_pred_nb = model.predict_proba(X_val_tf)\n",
    "print('mean ROC-AUC on validation set:', roc_auc_score(y_val, y_pred_nb))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- so what is OneVsRestClassifier doing here? \n",
    "\n",
    "- This is basically wrapper that converts the problem of 6 class multi-label problem into 6 individual binary classifications .i.e it builds 6 multinomialNB models. Let me write this implement what I said & verify my understanding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Naive Bayes\n",
      "\n",
      "Class:= toxic\n",
      "Train ROC AUC: 0.963033817358675\n",
      "Valid ROC AUC: 0.9433242109626461\n",
      "\n",
      "Class:= severe_toxic\n",
      "Train ROC AUC: 0.960087491016496\n",
      "Valid ROC AUC: 0.9384428174271573\n",
      "\n",
      "Class:= obscene\n",
      "Train ROC AUC: 0.9644609483402\n",
      "Valid ROC AUC: 0.9402753274549794\n",
      "\n",
      "Class:= threat\n",
      "Train ROC AUC: 0.9045076993073891\n",
      "Valid ROC AUC: 0.8043702907286571\n",
      "\n",
      "Class:= insult\n",
      "Train ROC AUC: 0.9607587688582031\n",
      "Valid ROC AUC: 0.9369958039957598\n",
      "\n",
      "Class:= identity_hate\n",
      "Train ROC AUC: 0.9153727553335663\n",
      "Valid ROC AUC: 0.8649606561769773\n",
      "\n",
      "mean column-wise ROC AUC on Train data:  0.9447035800357547\n",
      "mean column-wise ROC AUC on Val data: 0.9047281844576962\n"
     ]
    }
   ],
   "source": [
    "model = MultinomialNB()\n",
    "\n",
    "train_rocs = []\n",
    "valid_rocs = []\n",
    "\n",
    "preds_train = np.zeros(y_train.shape)\n",
    "preds_valid = np.zeros(y_val.shape)\n",
    "preds_test = np.zeros((len(test_df), len(label_col)))\n",
    "print('model: Naive Bayes')\n",
    "for i, label_name in enumerate(label_col):\n",
    "    print('\\nClass:= '+label_name)\n",
    "    # fit\n",
    "    model.fit(X_train_tf,y_train[label_name])\n",
    "    \n",
    "    # train\n",
    "    preds_train[:,i] = model.predict_proba(X_train_tf)[:,1]\n",
    "    train_roc_class = roc_auc_score(y_train[label_name],preds_train[:,i])\n",
    "    print('Train ROC AUC:', train_roc_class)\n",
    "    train_rocs.append(train_roc_class)\n",
    "\n",
    "    # valid\n",
    "    preds_valid[:,i] = model.predict_proba(X_val_tf)[:,1]\n",
    "    valid_roc_class = roc_auc_score(y_val[label_name],preds_valid[:,i])\n",
    "    print('Valid ROC AUC:', valid_roc_class)\n",
    "    valid_rocs.append(valid_roc_class)\n",
    "    \n",
    "    # test predictions\n",
    "    preds_test[:,i] = model.predict_proba(X_test_tf)[:,1]\n",
    "    \n",
    "print('\\nmean column-wise ROC AUC on Train data: ', np.mean(train_rocs))\n",
    "print('mean column-wise ROC AUC on Val data:', np.mean(valid_rocs))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- And the results of both implementations are excatly same..\n",
    "- Naive Bayes performed pretty well in its zone with 0.9047... \n",
    "- This looks like a good score only until we ecperiment with other models...as they might score a way higher lets see...how linear models like Logistic regression & linear SVM perform with default parameters\n",
    "- From here on I will be using the onevsrest classifier."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2. Logisitc Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Logistic Regression\n",
      "mean ROC-AUC on train set: 0.9916173033614236\n",
      "mean ROC-AUC on validation set: 0.9765313638921876\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "model = OneVsRestClassifier(LogisticRegression(), n_jobs = -1)\n",
    "model.fit(X_train_tf, y_train)\n",
    "print('model: Logistic Regression')\n",
    "print('mean ROC-AUC on train set:', roc_auc_score(y_train, model.predict_proba(X_train_tf)))\n",
    "y_pred_log = model.predict_proba(X_val_tf)\n",
    "print('mean ROC-AUC on validation set:', roc_auc_score(y_val, y_pred_log))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3. Linear SVM\n",
    "\n",
    "- Linear SVM will not support probability prediction as hinge loss is unstable, the way around is using a calibration classifer wrapper and it wont support OnevsRest classifers. so we I am using manual python loop again.\n",
    "\n",
    "- Notice that I fitted a new `CalibratedClassifierCV` wrapper afer fitting the `LinearSVC` classifier."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Linear SVM\n",
      "\n",
      "Class:= toxic\n",
      "Train ROC AUC: 0.9978703697448817\n",
      "Valid ROC AUC: 0.96089758073119\n",
      "\n",
      "Class:= severe_toxic\n",
      "Train ROC AUC: 0.9866976038380895\n",
      "Valid ROC AUC: 0.9840277745892984\n",
      "\n",
      "Class:= obscene\n",
      "Train ROC AUC: 0.9871142993892349\n",
      "Valid ROC AUC: 0.9766191922810586\n",
      "\n",
      "Class:= threat\n",
      "Train ROC AUC: 0.957937067118062\n",
      "Valid ROC AUC: 0.9371750021897756\n",
      "\n",
      "Class:= insult\n",
      "Train ROC AUC: 0.9816784291545445\n",
      "Valid ROC AUC: 0.9723417577232863\n",
      "\n",
      "Class:= identity_hate\n",
      "Train ROC AUC: 0.9618547477210576\n",
      "Valid ROC AUC: 0.9461706736620115\n",
      "\n",
      "mean column-wise ROC AUC on Train data:  0.978858752827645\n",
      "mean column-wise ROC AUC on Val data: 0.9628719968627699\n"
     ]
    }
   ],
   "source": [
    "from sklearn.svm import LinearSVC\n",
    "from sklearn.calibration import CalibratedClassifierCV\n",
    "\n",
    "model = LinearSVC()\n",
    "\n",
    "train_rocs = []\n",
    "valid_rocs = []\n",
    "\n",
    "preds_train = np.zeros(y_train.shape)\n",
    "preds_valid = np.zeros(y_val.shape)\n",
    "preds_test = np.zeros((len(test_df), len(label_col)))\n",
    "print('model: Linear SVM')\n",
    "for i, label_name in enumerate(label_col):\n",
    "    print('\\nClass:= '+label_name)\n",
    "    \n",
    "    # fit\n",
    "    model.fit(X_train_tf,y_train[label_name])\n",
    "    \n",
    "    # calibration classifier fit\n",
    "    model = CalibratedClassifierCV(model, cv = 'prefit')\n",
    "    model.fit(X_train_tf, y_train[label_name])\n",
    "    \n",
    "    # train\n",
    "    preds_train[:,i] = model.predict_proba(X_train_tf)[:,1]\n",
    "    train_roc_class = roc_auc_score(y_train[label_name],preds_train[:,i])\n",
    "    print('Train ROC AUC:', train_roc_class)\n",
    "    train_rocs.append(train_roc_class)\n",
    "\n",
    "    # valid\n",
    "    preds_valid[:,i] = model.predict_proba(X_val_tf)[:,1]\n",
    "    valid_roc_class = roc_auc_score(y_val[label_name],preds_valid[:,i])\n",
    "    print('Valid ROC AUC:', valid_roc_class)\n",
    "    valid_rocs.append(valid_roc_class)\n",
    "    \n",
    "    # test predictions\n",
    "    preds_test[:,i] = model.predict_proba(X_test_tf)[:,1]\n",
    "    \n",
    "print('\\nmean column-wise ROC AUC on Train data: ', np.mean(train_rocs))\n",
    "print('mean column-wise ROC AUC on Val data:', np.mean(valid_rocs))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Performance of Linear Models:\n",
    "\n",
    "- Both models Logistic & SVM are performing better than naive bayes models.\n",
    "- Although Logistic is better with default parameters, I need to experiment & see if fine tuning helps Linear SVM.\n",
    "- Moreover, these models are interpretable & not very complex in terms of time & space required.\n",
    "- Further with prerocessing & fine tuning of models I am hoping to reach ROC-AUC of 0.98.\n",
    "\n",
    "\n",
    "Now I will be moving to baselines of non-linear decision tree models namely RandomForests, XGB and Lightgbm.\n",
    "#### 4. RandomForests:\n",
    "\n",
    "why not single decision tree: \n",
    "\n",
    "- In my practice I found RandomForest always outperforms decision trees...in other words where ever decision trees work well..certainly randomforests will also do as its base learner is a decision tree.\n",
    "\n",
    "- Randomforests also outperforms as it averages the decision from 100's of decision tree models. Moreover, decision tree easily tends to overfit relying only on a single decision tree."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Random Forest\n",
      "mean ROC-AUC on train set: 0.9998282082745614\n",
      "mean ROC-AUC on validation set: 0.8719848045825899\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "model = OneVsRestClassifier(RandomForestClassifier(), n_jobs = -1)\n",
    "print('model: Random Forest')\n",
    "model.fit(X_train_tf, y_train)\n",
    "print('mean ROC-AUC on train set:', roc_auc_score(y_train, model.predict_proba(X_train_tf)))\n",
    "y_pred_rf = model.predict_proba(X_val_tf)\n",
    "print('mean ROC-AUC on validation set:', roc_auc_score(y_val, y_pred_rf))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5. LightGBM\n",
    "\n",
    "- optimized version and generally takes lesser time than XGBoost."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Lightgbm\n",
      "mean ROC-AUC on train set: 0.966037582982347\n",
      "mean ROC-AUC on validation set: 0.9240550756463076\n"
     ]
    }
   ],
   "source": [
    "from lightgbm import LGBMClassifier\n",
    "model = OneVsRestClassifier(LGBMClassifier(), n_jobs = -1)\n",
    "print('model: Lightgbm')\n",
    "model.fit(X_train_tf, y_train)\n",
    "print('mean ROC-AUC on train set:', roc_auc_score(y_train, model.predict_proba(X_train_tf)))\n",
    "y_pred_log = model.predict_proba(X_val_tf)\n",
    "print('mean ROC-AUC on validation set:', roc_auc_score(y_val, y_pred_log))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 6.XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: XGBoost\n",
      "mean ROC-AUC on train set: 0.9498677089044373\n",
      "mean ROC-AUC on validation set: 0.9366437176804424\n"
     ]
    }
   ],
   "source": [
    "from xgboost import XGBClassifier\n",
    "model = OneVsRestClassifier(XGBClassifier(), n_jobs = -1)\n",
    "print('model: XGBoost')\n",
    "model.fit(X_train_tf, y_train)\n",
    "print('mean ROC-AUC on train set:', roc_auc_score(y_train, model.predict_proba(X_train_tf)))\n",
    "y_pred_lgb = model.predict_proba(X_val_tf)\n",
    "print('mean ROC-AUC on validation set:', roc_auc_score(y_val, y_pred_lgb))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Random Forest are serverly overfitting with train ROC-AUC at 0.99 and validation at 0.87\n",
    "- All these non-linear models might be failing case of classes like 'threat' with only around 500 samples for this class which could be a reasoning - bagging & boosting requires more data ?\n",
    "- Non-linear models in this case are overfitting a bit, when compared to linear models.\n",
    "- Guessing, Non linear models can be improved by reducing features using selecting only Top K best or by limiting max_depth, increasing num_trees and tuning other hyperparameters . But ensmebles are also more time complex so sticking with simple linear models is better in this case."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Conclusions:\n",
    "- Linear models are very well suited for this problem\n",
    "- Logistic Regression is having better baseline score than any other model\n",
    "- Logistic also has an edge with lower time/space complexity compared to ensembles and also, we can impove this further by text preprocessing & hyper parameter tuning.\n",
    "- So rather than trying complex models, I will settle with linear models & will try to improve the performance using preprocessing, finetuning & feature engineering. My next target is to improve to score to more than 0.98 using only linear models.\n",
    "\n",
    "### how about using deep neural nets, CNNs & LSTMs ?\n",
    "- with amount of data we are having it is possible for deep neural nets might provide better results.\n",
    "- But Let say after preprocessig & tuning our final Logistic model is having ROC-AUC of 0.98 on average & for some LSTM model is having ROC-AUC of 0.99..I am question myself, is it worth training & deploying such complex model which we can't interpret for an improvement of just 0.01 score ? I would say **NO** but in the kaggle competitions, every improvement in 0.0001 score matters...so may be we can try LSTMs after all to improve our knowledge..but in production I would prefer to say NO Again it ultimately depends on serverity of the problem & several other factors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Logistic Regression\n",
      "mean ROC-AUC on train set: 0.991368050986793\n"
     ]
    }
   ],
   "source": [
    "X_train_val = train_df['comment_text']\n",
    "y_train_val = train_df[label_col]\n",
    "\n",
    "X_train_val = tfidf.fit_transform(train_df['comment_text'])\n",
    "X_test = tfidf.transform(test_df['comment_text'])\n",
    "\n",
    "\n",
    "model = OneVsRestClassifier(LogisticRegression(), n_jobs = -1)\n",
    "model.fit(X_train_val, y_train_val)\n",
    "print('model: Logistic Regression')\n",
    "print('mean ROC-AUC on train set:', roc_auc_score(y_train_val, model.predict_proba(X_train_val)))\n",
    "y_test_pred = model.predict_proba(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<a href='submission.csv' target='_blank'>submission.csv</a><br>"
      ],
      "text/plain": [
       "/kaggle/working/submission.csv"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## making a submission file\n",
    "sub_df.iloc[:,1:] = y_test_pred\n",
    "sub_df.head()\n",
    "from IPython.display import FileLink\n",
    "sub_df.to_csv('submission.csv', index = None)\n",
    "FileLink('submission.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "_kg_hide-input": true
   },
   "outputs": [],
   "source": [
    "# sorted_features = sorted(zip(model.coef_.ravel(), feature_names))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "_kg_hide-input": true
   },
   "outputs": [],
   "source": [
    "#"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}