Classification Notebook.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Amazon Video Game Rating - Classification vs Regression."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We extract data from json files. Description of files and preprocessing described in detail in attached paper.\n",
    "This notebook explores various models, does preprocessing and looks at dataset analysis. To see Bag of Words models and implementations with preprocessing, please see other attached notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading data...\n",
      "done\n"
     ]
    }
   ],
   "source": [
    "import numpy\n",
    "from urllib.request import urlopen\n",
    "import scipy.optimize\n",
    "import random\n",
    "import matplotlib.pyplot as plt\n",
    "plt.style.use('seaborn-whitegrid')\n",
    "import gzip\n",
    "import json\n",
    "\n",
    "def parseDataFromURL(fname):\n",
    "  for l in urlopen(fname):\n",
    "    yield eval(l)\n",
    "\n",
    "def parseData(fname):\n",
    "  for l in open(fname):\n",
    "    yield eval(l)\n",
    "    \n",
    "def parse(path):\n",
    "    g = gzip.open(path, 'r')\n",
    "    for l in g:\n",
    "        yield json.loads(l)\n",
    "        \n",
    "    \n",
    "print(\"Reading data...\")\n",
    "# Download from http://jmcauley.ucsd.edu/cse255/data/beer/beer_50000.json\"\n",
    "data = list(parse(\"Downloads/Video_Games_5.json.gz\"))\n",
    "print(\"done\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "497577"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'overall': 1.0,\n",
       " 'vote': '28',\n",
       " 'verified': False,\n",
       " 'reviewTime': '05 15, 2012',\n",
       " 'reviewerID': 'A1LMJ9W8UX1H5B',\n",
       " 'asin': '0700026657',\n",
       " 'reviewerName': 'Rob NY',\n",
       " 'reviewText': \"I've bought and played ALL of the ANNO games since 1602, and spent more hours than I can even count building massive empires with extremely complex trade routes. I was thrilled to see another installment released, but I absolutely will NOT pay for something that limits my hardware changes and installs.\\n\\nWake up UBI, look at all of your other software title reviews, both here on Amazon and on gaming sites. People are not as stupid as you think, and will simply find a different game to fill their time with and throw their money at if their freedoms are inhibited by your software.\\n\\nBy adding DRM you've actually promoted the use of piracy instead of reduced it, because those that really do want to play it will find a way.\",\n",
       " 'summary': 'SAY NO TO DRM!!',\n",
       " 'unixReviewTime': 1337040000}"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data[11]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Introducting a baseline accuracy. Trivial case"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6024374116970841"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "##Baseline accuracy by predicting mode of the labelss\n",
    "(len([d for d in data if d['overall'] == 5]))/len(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "##seperate into validation training set\n",
    "random.shuffle(data)\n",
    "training = data[:487577]\n",
    "validation = data[487577:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import defaultdict"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Creating a similarity function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "usersPerItem = defaultdict(set)\n",
    "itemsPerUser = defaultdict(set)\n",
    "reviewsPerUser = defaultdict(list)\n",
    "reviewsPerItem = defaultdict(list)\n",
    "for d in training:\n",
    "    user,item = d['reviewerID'], d['asin']\n",
    "    usersPerItem[item].add(user)\n",
    "    itemsPerUser[user].add(item)\n",
    "    reviewsPerUser[user].append(d)\n",
    "    reviewsPerItem[item].append(d)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'0700026657', 'B005TMFXZQ', 'B00657AVMO', 'B00684AGQ0', 'B00KWF38AW'}"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "itemsPerUser['A1LMJ9W8UX1H5B']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "import statistics\n",
    "ratingMean = statistics.mean([d['overall'] for d in training])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "ratingMode = statistics.mode([d['overall'] for d in training])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d['vote'] for d in training if 'vote' in d and d['vote'] == 0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "## implement similarity based model\n",
    "def predictRating(user,item):\n",
    "    ratings = []\n",
    "    similarities = []\n",
    "    for d in reviewsPerUser[user]:\n",
    "        i2 = d['asin']\n",
    "        if i2 == item: continue\n",
    "        ratings.append((d['overall']))\n",
    "        similarities.append(Jaccard(usersPerItem[item],usersPerItem[i2]))\n",
    "    if (sum(similarities) > 0):\n",
    "        weightedRatings = [(x*y) for x,y in zip(ratings,similarities)]\n",
    "        return ((sum(weightedRatings) / sum(similarities)))\n",
    "    else:\n",
    "        # User hasn't rated any similar items\n",
    "        return ratingMode"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "def Jaccard(s1, s2):\n",
    "    numer = len(s1.intersection(s2))\n",
    "    denom = len(s1.union(s2))\n",
    "    return numer / denom"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'overall': 5.0,\n",
       " 'verified': True,\n",
       " 'reviewTime': '06 8, 2015',\n",
       " 'reviewerID': 'A2860P952GFG39',\n",
       " 'asin': 'B00NP8J2LC',\n",
       " 'style': {'Format:': ' Video Game'},\n",
       " 'reviewerName': 'Cougar Gold',\n",
       " 'reviewText': 'Happy with the gift.',\n",
       " 'summary': 'Five Stars',\n",
       " 'unixReviewTime': 1433721600}"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "training[14]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictRating('A3GBULTIVS4LNA', 'B00NAW5UZA')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6006"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "##always predict mode\n",
    "len([d['overall'] for d in validation if d['overall'] == 5])/len(validation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5472"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "##accuracy\n",
    "correct = 0\n",
    "for entry in validation:\n",
    "    if entry['overall'] == predictRating(entry['reviewerID'], entry['asin']):\n",
    "        correct += 1\n",
    "        \n",
    "correct/len(validation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.4143744036022714\n"
     ]
    }
   ],
   "source": [
    "sums = []\n",
    "for d in validation:\n",
    "    sums.append((d['overall'] - ratingMean) * (d['overall'] - ratingMean))\n",
    "sums = sum(sums)/len(sums)\n",
    "print(sums)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.5362\n"
     ]
    }
   ],
   "source": [
    "sums = []\n",
    "for d in validation:\n",
    "    sums.append((d['overall'] - predictRating(d['reviewerID'], d['asin'])) * (d['overall'] - predictRating(d['reviewerID'], d['asin'])))\n",
    "sums = sum(sums)/len(sums)\n",
    "print(sums)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# User - User based similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "##trying user user similarity\n",
    "\n",
    "def predictRating(user,item):\n",
    "    \n",
    "    ratings = []\n",
    "    similarities = []\n",
    "    for d in reviewsPerItem[item]:\n",
    "        u2 = d['asin']\n",
    "        if u2 == user: continue\n",
    "        ratings.append(d['overall'])\n",
    "        similarities.append(Jaccard(itemsPerUser[user],itemsPerUser[u2]))\n",
    "    if (sum(similarities) > 0):\n",
    "        weightedRatings = [(x*y) for x,y in zip(ratings,similarities)]\n",
    "        #print(\"if\")\n",
    "        return round(sum(weightedRatings) / sum(similarities))\n",
    "    else:\n",
    "        # User hasn't rated any similar items\n",
    "        #print('else')\n",
    "        return ratingMode"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6044"
      ]
     },
     "execution_count": 99,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "##accuracy\n",
    "\n",
    "correct = 0\n",
    "for entry in validation:\n",
    "    if entry['overall'] == predictRating(entry['reviewerID'], entry['asin']):\n",
    "        correct += 1\n",
    "        \n",
    "correct/len(validation)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [],
   "source": [
    "## implement similarity based model\n",
    "def predictRating(user,item):\n",
    "    ratings = []\n",
    "    similarities = []\n",
    "    for d in reviewsPerUser[user]:\n",
    "        i2 = d['asin']\n",
    "        if i2 == item: continue\n",
    "        ratings.append(d['overall'])\n",
    "        similarities.append(Jaccard(usersPerItem[item],usersPerItem[i2]))\n",
    "    if (sum(similarities) > 0):\n",
    "        weightedRatings = [(x*y) for x,y in zip(ratings,similarities)]\n",
    "        #print('if')\n",
    "        return round(sum(weightedRatings) / sum(similarities))\n",
    "    else:\n",
    "        # User hasn't rated any similar items\n",
    "        #print('else')\n",
    "        return ratingMode"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.5472\n"
     ]
    }
   ],
   "source": [
    "correct = 0\n",
    "for entry in validation:\n",
    "    if entry['overall'] == predictRating(entry['reviewerID'], entry['asin']):\n",
    "        correct += 1\n",
    "        \n",
    "print(correct/len(validation))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Introducing thresholds / Pipeling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [],
   "source": [
    "##trying deviation from a threshold\n",
    "def predictRating(user,item):\n",
    "    dev = ratingMean\n",
    "    ratings = []\n",
    "    similarities = []\n",
    "    for d in reviewsPerUser[user]:\n",
    "        i2 = d['asin']\n",
    "        if i2 == item: continue\n",
    "        ratings.append(d['overall'] - dev)\n",
    "        similarities.append(Jaccard(usersPerItem[item],usersPerItem[i2]))\n",
    "    if (sum(similarities) > 0):\n",
    "        weightedRatings = [(x*y) for x,y in zip(ratings,similarities)]\n",
    "        #print('if')\n",
    "        #print((sum(weightedRatings) / sum(similarities)))\n",
    "        returnval= (dev + ((sum(weightedRatings) / sum(similarities)) ))\n",
    "        return returnval  \n",
    "    else:\n",
    "        # User hasn't rated any similar items\n",
    "        #print('else')\n",
    "        return ratingMean"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.266\n"
     ]
    }
   ],
   "source": [
    "correct = 0\n",
    "for entry in validation:\n",
    "    if entry['overall'] == predictRating(entry['reviewerID'], entry['asin']):\n",
    "        correct += 1\n",
    "        \n",
    "print(correct/len(validation))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.3749404173699182\n"
     ]
    }
   ],
   "source": [
    "##MSE\n",
    "sums = []\n",
    "for d in validation:\n",
    "    sums.append((d['overall'] - predictRating(d['reviewerID'], d['asin'])) * (d['overall'] - predictRating(d['reviewerID'], d['asin'])))\n",
    "sums = sum(sums)/len(sums)\n",
    "print(sums)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [],
   "source": [
    "##Similarity is not working\n",
    "def predictRating(user,item):\n",
    "    dev = ratingMean\n",
    "    ratings = []\n",
    "    similarities = []\n",
    "    for d in reviewsPerUser[user]:\n",
    "        i2 = d['asin']\n",
    "        if i2 == item: continue\n",
    "        ratings.append(d['overall'] - dev)\n",
    "        similarities.append(Jaccard(usersPerItem[item],usersPerItem[i2]))\n",
    "    if (sum(similarities) > 0 and len(similarities) > 6):\n",
    "        weightedRatings = [(x*y) for x,y in zip(ratings,similarities)]\n",
    "        #print('if')\n",
    "        #print((sum(weightedRatings) / sum(similarities)))\n",
    "        returnval = dev + (((sum(weightedRatings) / sum(similarities)) ))\n",
    "        return returnval  \n",
    "    else:\n",
    "        # User hasn't rated any similar items\n",
    "        #print('else')\n",
    "        return ratingMean"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.1393\n"
     ]
    }
   ],
   "source": [
    "correct = 0\n",
    "for entry in validation:\n",
    "    if entry['overall'] == predictRating(entry['reviewerID'], entry['asin']):\n",
    "        correct += 1\n",
    "        \n",
    "print(correct/len(validation))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.3181344193377496\n"
     ]
    }
   ],
   "source": [
    "##MSE\n",
    "sums = []\n",
    "for d in validation:\n",
    "    sums.append((d['overall'] - predictRating(d['reviewerID'], d['asin'])) * (d['overall'] - predictRating(d['reviewerID'], d['asin'])))\n",
    "sums = sum(sums)/len(sums)\n",
    "print(sums)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'overall': 4.0,\n",
       " 'verified': True,\n",
       " 'reviewTime': '01 9, 2015',\n",
       " 'reviewerID': 'A1PL8WEPH9H7N8',\n",
       " 'asin': 'B00BWBTJ4Y',\n",
       " 'reviewerName': 'Matthew Opsahl',\n",
       " 'reviewText': 'It is nice to get different colors so everyone can know whose controller is whose',\n",
       " 'summary': 'Four Stars',\n",
       " 'unixReviewTime': 1420761600}"
      ]
     },
     "execution_count": 138,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "training[2]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SVM with simple features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We notice poor accuracy because of the nature of dataset. Points cannot be linearly separable with simple feature\n",
    "representations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.svm import SVC\n",
    "from sklearn import svm, datasets\n",
    "from sklearn.metrics import accuracy_score\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "okrunning\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "SVC()"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "##support vector machine\n",
    "\n",
    "\n",
    "y_train = []\n",
    "'''\n",
    "for d in training:\n",
    "    newvec = ohe.copy()\n",
    "    if(int(d['overall']) > 1 ):\n",
    "        newvec[int(d['overall']) - 2] = 1\n",
    "    y_train.append(newvec)\n",
    "'''  \n",
    "print(\"okrunning\")\n",
    "y_train = [d['overall'] for d in training]\n",
    "x_train = [[len(r['reviewText']) if 'reviewText' in r else 0, r['verified'], int(r['vote'].replace(',','')) if 'vote' in r else 0] for r in training ]\n",
    "SupportVectorClassModel = SVC()\n",
    "SupportVectorClassModel.fit(x_train[:10000],y_train[:10000])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_val = [d['overall'] for d in validation]\n",
    "x_val = [[len(r['reviewText']) if 'reviewText' in r else 0, r['verified'], int(r['vote'].replace(',','')) if 'vote' in r else 0] for r in validation ]\n",
    "\n",
    "\n",
    "y_pred = SupportVectorClassModel.predict(x_train[:10000])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5992"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_val,y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "rbf = svm.SVC(kernel='rbf', gamma=1, C=0.1).fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "rbf_pred = rbf.predict(x_val[:10000])\n",
    "rbf_accuracy = accuracy_score(y_val, rbf_pred)\n",
    "print(rbf_accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_val = [d['overall'] for d in validation]\n",
    "\n",
    "x_val = [[len(r['reviewText']), r['verified']] if 'reviewText' in r else [0, r['verified']] for r in validation ]\n",
    "\n",
    "y_pred = SupportVectorClassModel.predict(x_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6044"
      ]
     },
     "execution_count": 163,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_val,y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6044"
      ]
     },
     "execution_count": 195,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d['overall'] for d in validation if d['overall'] == 5])/len(validation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "6027"
      ]
     },
     "execution_count": 204,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d['overall'] for d in training[:10000] if d['overall'] == 5 ])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy\n",
    "from urllib.request import urlopen\n",
    "import scipy.optimize\n",
    "import random\n",
    "from sklearn import linear_model\n",
    "from sklearn.ensemble import RandomForestClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "#ohe = [0, 0, 0, 0]\n",
    "x_train = []\n",
    "y_train = []\n",
    "\n",
    "y_train = [d['overall'] for d in training]\n",
    "x_train = [[len(r['reviewText']) if 'reviewText' in r else 0, r['verified'], int(r['vote'].replace(',','')) if 'vote' in r else 0] for r in training]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "mod = linear_model.LogisticRegression(C=1.0, class_weight = 'balanced', max_iter = 1000, solver='lbfgs')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(class_weight='balanced', max_iter=1000)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mod.fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_val = [d['overall'] for d in validation]\n",
    "x_val = x_train = [[len(r['reviewText']) if 'reviewText' in r else 0, r['verified'], int(r['vote'].replace(',','')) if 'vote' in r else 0] for r in validation]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = mod.predict(x_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.4814\n"
     ]
    }
   ],
   "source": [
    "#Accuracy on the validation set\n",
    "correct = 0\n",
    "for i in range(len(validation)):\n",
    "    if(y_pred[i] == y_val[i]):\n",
    "        correct += 1\n",
    "print(correct/len(validation))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# RandomForestClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "rfc = RandomForestClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = [d['overall'] for d in training]\n",
    "x_train = [[len(r['reviewText']) if 'reviewText' in r else 0, r['verified'], int(r['vote'].replace(',','')) if 'vote' in r else 0] for r in training]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier()"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rfc.fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_val = [d['overall'] for d in validation]\n",
    "x_val = x_train = [[len(r['reviewText']) if 'reviewText' in r else 0, r['verified'], int(r['vote'].replace(',','')) if 'vote' in r else 0] for r in validation]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "rfcpred = rfc.predict(x_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.595\n"
     ]
    }
   ],
   "source": [
    "#Accuracy on the validation set\n",
    "correct = 0\n",
    "for i in range(len(validation)):\n",
    "    if(rfcpred[i] == y_val[i]):\n",
    "        correct += 1\n",
    "print(correct/len(validation))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Set Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = [len(r['reviewText']) if 'reviewText' in r else 0 for r in data]\n",
    "y = [r['overall'] for r in data]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(y, x)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = [int(r['vote'].replace(',','')) if 'vote' in r else 0 for r in data]\n",
    "y = [r['overall'] for r in data]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(y, x)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = [r['unixReviewTime'] for r in data]\n",
    "y = [r['overall'] for r in data]\n",
    "plt.scatter(y, x)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# NLP models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from adaptnlp import EasySequenceClassifier\n",
    "\n",
    "classifier = EasySequenceClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'overall': 4.0,\n",
       " 'verified': False,\n",
       " 'reviewTime': '07 27, 2015',\n",
       " 'reviewerID': 'A1JGAP0185YJI6',\n",
       " 'asin': '0700026657',\n",
       " 'reviewerName': 'travis',\n",
       " 'reviewText': 'I played it a while but it was alright. The steam was a bit of trouble. The more they move these game to steam the more of a hard time I have activating and playing a game. But in spite of that it was fun, I liked it. Now I am looking forward to anno 2205 I really want to play my way to the moon.',\n",
       " 'summary': 'But in spite of that it was fun, I liked it',\n",
       " 'unixReviewTime': 1437955200}"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_json(r\"Downloads/Video_Games_5.json\", lines = True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.drop('image', axis = 1)\n",
    "df = df.drop('style', axis = 1)\n",
    "df = df.drop('vote', axis = 1)\n",
    "df = df.drop('unixReviewTime', axis = 1)\n",
    "df = df.drop('summary', axis = 1)\n",
    "df = df.drop('reviewTime', axis = 1)\n",
    "df = df.drop('verified', axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>overall</th>\n",
       "      <th>reviewerID</th>\n",
       "      <th>asin</th>\n",
       "      <th>reviewerName</th>\n",
       "      <th>reviewText</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5</td>\n",
       "      <td>A1HP7NVNPFMA4N</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Ambrosia075</td>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4</td>\n",
       "      <td>A1JGAP0185YJI6</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>travis</td>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>A1YJWEXHQBWK2B</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Vincent G. Mezera</td>\n",
       "      <td>ok game.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2</td>\n",
       "      <td>A2204E1TH211HT</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Grandma KR</td>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>A2RF5B5H74JLPE</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>jon</td>\n",
       "      <td>great game, I love it and have played it since...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497572</th>\n",
       "      <td>4</td>\n",
       "      <td>AVECM71LSZLC5</td>\n",
       "      <td>B01HGPUTCA</td>\n",
       "      <td>boris teplitskiy</td>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497573</th>\n",
       "      <td>3</td>\n",
       "      <td>A1RS06313BL6WN</td>\n",
       "      <td>B01HH6JEOC</td>\n",
       "      <td>Tom Stopsign</td>\n",
       "      <td>Okay stuff.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497574</th>\n",
       "      <td>3</td>\n",
       "      <td>ACIZ77IGIX2JL</td>\n",
       "      <td>B01HH6JEOC</td>\n",
       "      <td>Era</td>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497575</th>\n",
       "      <td>4</td>\n",
       "      <td>A34GG58TJ1A3SH</td>\n",
       "      <td>B01HIZF7XE</td>\n",
       "      <td>seamonkey10</td>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497576</th>\n",
       "      <td>2</td>\n",
       "      <td>A6W81WTFK940B</td>\n",
       "      <td>B01HIZGKOE</td>\n",
       "      <td>msam420</td>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497577 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        overall      reviewerID        asin       reviewerName  \\\n",
       "0             5  A1HP7NVNPFMA4N  0700026657        Ambrosia075   \n",
       "1             4  A1JGAP0185YJI6  0700026657             travis   \n",
       "2             3  A1YJWEXHQBWK2B  0700026657  Vincent G. Mezera   \n",
       "3             2  A2204E1TH211HT  0700026657         Grandma KR   \n",
       "4             5  A2RF5B5H74JLPE  0700026657                jon   \n",
       "...         ...             ...         ...                ...   \n",
       "497572        4   AVECM71LSZLC5  B01HGPUTCA   boris teplitskiy   \n",
       "497573        3  A1RS06313BL6WN  B01HH6JEOC       Tom Stopsign   \n",
       "497574        3   ACIZ77IGIX2JL  B01HH6JEOC                Era   \n",
       "497575        4  A34GG58TJ1A3SH  B01HIZF7XE        seamonkey10   \n",
       "497576        2   A6W81WTFK940B  B01HIZGKOE            msam420   \n",
       "\n",
       "                                               reviewText  \n",
       "0       This game is a bit hard to get the hang of, bu...  \n",
       "1       I played it a while but it was alright. The st...  \n",
       "2                                                ok game.  \n",
       "3       found the game a bit too complicated, not what...  \n",
       "4       great game, I love it and have played it since...  \n",
       "...                                                   ...  \n",
       "497572                 not OEM but good replacement parts  \n",
       "497573                                        Okay stuff.  \n",
       "497574  This does add some kids room things that are v...  \n",
       "497575  I think I originally began playing Bioshock se...  \n",
       "497576  The graphics are terrible, it looks like ps2 g...  \n",
       "\n",
       "[497577 rows x 5 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.dropna(axis = 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>overall</th>\n",
       "      <th>reviewerID</th>\n",
       "      <th>asin</th>\n",
       "      <th>reviewerName</th>\n",
       "      <th>reviewText</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5</td>\n",
       "      <td>A1HP7NVNPFMA4N</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Ambrosia075</td>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4</td>\n",
       "      <td>A1JGAP0185YJI6</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>travis</td>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>A1YJWEXHQBWK2B</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Vincent G. Mezera</td>\n",
       "      <td>ok game.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2</td>\n",
       "      <td>A2204E1TH211HT</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Grandma KR</td>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>A2RF5B5H74JLPE</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>jon</td>\n",
       "      <td>great game, I love it and have played it since...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497572</th>\n",
       "      <td>4</td>\n",
       "      <td>AVECM71LSZLC5</td>\n",
       "      <td>B01HGPUTCA</td>\n",
       "      <td>boris teplitskiy</td>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497573</th>\n",
       "      <td>3</td>\n",
       "      <td>A1RS06313BL6WN</td>\n",
       "      <td>B01HH6JEOC</td>\n",
       "      <td>Tom Stopsign</td>\n",
       "      <td>Okay stuff.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497574</th>\n",
       "      <td>3</td>\n",
       "      <td>ACIZ77IGIX2JL</td>\n",
       "      <td>B01HH6JEOC</td>\n",
       "      <td>Era</td>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497575</th>\n",
       "      <td>4</td>\n",
       "      <td>A34GG58TJ1A3SH</td>\n",
       "      <td>B01HIZF7XE</td>\n",
       "      <td>seamonkey10</td>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497576</th>\n",
       "      <td>2</td>\n",
       "      <td>A6W81WTFK940B</td>\n",
       "      <td>B01HIZGKOE</td>\n",
       "      <td>msam420</td>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497343 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        overall      reviewerID        asin       reviewerName  \\\n",
       "0             5  A1HP7NVNPFMA4N  0700026657        Ambrosia075   \n",
       "1             4  A1JGAP0185YJI6  0700026657             travis   \n",
       "2             3  A1YJWEXHQBWK2B  0700026657  Vincent G. Mezera   \n",
       "3             2  A2204E1TH211HT  0700026657         Grandma KR   \n",
       "4             5  A2RF5B5H74JLPE  0700026657                jon   \n",
       "...         ...             ...         ...                ...   \n",
       "497572        4   AVECM71LSZLC5  B01HGPUTCA   boris teplitskiy   \n",
       "497573        3  A1RS06313BL6WN  B01HH6JEOC       Tom Stopsign   \n",
       "497574        3   ACIZ77IGIX2JL  B01HH6JEOC                Era   \n",
       "497575        4  A34GG58TJ1A3SH  B01HIZF7XE        seamonkey10   \n",
       "497576        2   A6W81WTFK940B  B01HIZGKOE            msam420   \n",
       "\n",
       "                                               reviewText  \n",
       "0       This game is a bit hard to get the hang of, bu...  \n",
       "1       I played it a while but it was alright. The st...  \n",
       "2                                                ok game.  \n",
       "3       found the game a bit too complicated, not what...  \n",
       "4       great game, I love it and have played it since...  \n",
       "...                                                   ...  \n",
       "497572                 not OEM but good replacement parts  \n",
       "497573                                        Okay stuff.  \n",
       "497574  This does add some kids room things that are v...  \n",
       "497575  I think I originally began playing Bioshock se...  \n",
       "497576  The graphics are terrible, it looks like ps2 g...  \n",
       "\n",
       "[497343 rows x 5 columns]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv (r\"Downloads/Video_Games_5csv.csv\", index = None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/transformers/training_args.py:347: FutureWarning: The `evaluate_during_training` argument is deprecated in favor of `evaluation_strategy` (which has more options)\n",
      "  FutureWarning,\n"
     ]
    }
   ],
   "source": [
    "from transformers import TrainingArguments\n",
    "\n",
    " \n",
    "training_args = TrainingArguments(\n",
    "\n",
    "output_dir='./models',\n",
    "\n",
    "num_train_epochs=1,\n",
    "\n",
    "per_device_train_batch_size=16,\n",
    "\n",
    "per_device_eval_batch_size=16,\n",
    "\n",
    "warmup_steps=500,\n",
    "\n",
    "weight_decay=0.01,\n",
    "\n",
    "evaluate_during_training=True,\n",
    "\n",
    "logging_dir='./logs',\n",
    "\n",
    "save_steps=100\n",
    "\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "77985b3512894c4c925685279d3679cd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=1817.0, style=ProgressStyle(description…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fd55d9e0ea604f4d8e8378dc2d32aa06",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=1227.0, style=ProgressStyle(description…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using custom data configuration default\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Downloading and preparing dataset ag_news/default (download: 29.88 MiB, generated: 30.23 MiB, post-processed: Unknown size, total: 60.10 MiB) to /Users/sri/.cache/huggingface/datasets/ag_news/default/0.0.0/fb5c5e74a110037311ef5e904583ce9f8b9fbc1354290f97b4929f01b3f48b1a...\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ee8a0e132bca41f88934c015fce53835",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=11045148.0, style=ProgressStyle(descrip…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "64951a01f6674e8aa6d3fb51c0bc7aa8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=751209.0, style=ProgressStyle(descripti…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9537cae92bdc4b7a8cf040cb90d8fa2c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=1.0, bar_style='info', layout=Layout(width='20px'), max=1.0), HTML(value=''…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7ca6e90a070541ff9ce91e862828827c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=1.0, bar_style='info', layout=Layout(width='20px'), max=1.0), HTML(value=''…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset ag_news downloaded and prepared to /Users/sri/.cache/huggingface/datasets/ag_news/default/0.0.0/fb5c5e74a110037311ef5e904583ce9f8b9fbc1354290f97b4929f01b3f48b1a. Subsequent calls will reuse this data.\n"
     ]
    }
   ],
   "source": [
    "from datasets import load_dataset\n",
    "\n",
    "train_dataset, eval_dataset = load_dataset('ag_news', split=['train[:10%]', 'test'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "49553423faca4d5db10d6d3576027564",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=433.0, style=ProgressStyle(description_…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cd587aefc6434f5a892e9c54b4291697",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=1.0, bar_style='info', description='Downloading', layout=Layout(width='20px…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9972eed811ba4582afb4738e8c953a5e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=1.0, bar_style='info', description='Downloading', layout=Layout(width='20px…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6f89fbd18b144f1e92537ae3bb2b4ba0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=435779157.0, style=ProgressStyle(descri…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-base-cased were not used when initializing BertForSequenceClassification: ['cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.seq_relationship.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.LayerNorm.bias']\n",
      "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-cased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    },
    {
     "ename": "ValueError",
     "evalue": "Must pass in `label_names` parameter for training when loading in CSV datasets.",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m--------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0mTraceback (most recent call last)",
      "\u001b[0;32m<ipython-input-19-26795bf3043b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     11\u001b[0m \u001b[0mtext_col_nm\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"reviewText\"\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     12\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 13\u001b[0;31m \u001b[0mlabel_col_nm\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"overall\"\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     14\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     15\u001b[0m )\n",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/adaptnlp/sequence_classification.py\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(self, training_args, train_dataset, eval_dataset, model_name_or_path, text_col_nm, label_col_nm, label_names)\u001b[0m\n\u001b[1;32m    488\u001b[0m                 \u001b[0mdata_path\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtrain_dataset\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    489\u001b[0m                 \u001b[0mlabel_col_nm\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mlabel_col_nm\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 490\u001b[0;31m                 \u001b[0mlabel_names\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mlabel_names\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    491\u001b[0m             )\n\u001b[1;32m    492\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0meval_dataset\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdatasets\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDataset\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/adaptnlp/sequence_classification.py\u001b[0m in \u001b[0;36m_csv2dataset\u001b[0;34m(self, data_path, label_col_nm, label_names)\u001b[0m\n\u001b[1;32m    549\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mlabel_names\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    550\u001b[0m             raise ValueError(\n\u001b[0;32m--> 551\u001b[0;31m                 \u001b[0;34m\"Must pass in `label_names` parameter for training when loading in CSV datasets.\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    552\u001b[0m             )\n\u001b[1;32m    553\u001b[0m         class_label = datasets.ClassLabel(\n",
      "\u001b[0;31mValueError\u001b[0m: Must pass in `label_names` parameter for training when loading in CSV datasets."
     ]
    }
   ],
   "source": [
    "classifier.train(\n",
    "\n",
    "training_args=training_args,\n",
    "\n",
    "train_dataset='Downloads/Video_Games_5csv.csv',\n",
    "\n",
    "eval_dataset='Downloads/Video_Games_5csv.csv',\n",
    "\n",
    "model_name_or_path=\"bert-base-cased\",\n",
    "\n",
    "text_col_nm=\"reviewText\",\n",
    "\n",
    "label_col_nm=\"overall\",\n",
    "\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/IPython/core/interactiveshell.py:3146: DtypeWarning: Columns (0) have mixed types.Specify dtype option on import or set low_memory=False.\n",
      "  interactivity=interactivity, compiler=compiler, result=result)\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "#data_path = \"data/\" \n",
    "data_file_address = 'Downloads/Video_Games_5csv.csv'\n",
    "\n",
    "df_data = pd.read_csv(data_file_address,sep=\",\",encoding=\"utf-8\",names=['overall','reviewText'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'Format:': ' Video Game'}                                       129039\n",
       "{'Platform:': ' PlayStation 3'}                                   10656\n",
       "{'Platform:': ' Xbox 360'}                                        10639\n",
       "{'Edition:': ' Standard', 'Platform:': ' PlayStation 4'}           9219\n",
       "{'Platform:': ' PC'}                                               8424\n",
       "                                                                  ...  \n",
       "{'Color:': ' Red Mechanical Switch', 'Style:': ' Excalibur'}          1\n",
       "{'Size:': ' Laser', 'Style:': ' GMS7011'}                             1\n",
       "{'Color:': ' Black & White'}                                          1\n",
       "{'Color:': ' Blue Mechanical Switch', 'Style:': ' Excalibur'}         1\n",
       "{'Color:': ' Forza 2'}                                                1\n",
       "Name: overall, Length: 2324, dtype: int64"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Have a look at the dataset\n",
    "df_data.head(n=20)\n",
    "\n",
    "# Have a look labels \n",
    "df_data.overall.unique()\n",
    "\n",
    "# Analyse the labels distribution\n",
    "df_data.overall.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# XLNet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/IPython/core/interactiveshell.py:3146: DtypeWarning: Columns (0) have mixed types.Specify dtype option on import or set low_memory=False.\n",
      "  interactivity=interactivity, compiler=compiler, result=result)\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "\n",
    "train_df = pd.read_csv('Downloads/Video_Games_5csv.csv', header=None)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>overall</td>\n",
       "      <td>reviewerID</td>\n",
       "      <td>asin</td>\n",
       "      <td>reviewerName</td>\n",
       "      <td>reviewText</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5</td>\n",
       "      <td>A1HP7NVNPFMA4N</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Ambrosia075</td>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4</td>\n",
       "      <td>A1JGAP0185YJI6</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>travis</td>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>A1YJWEXHQBWK2B</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Vincent G. Mezera</td>\n",
       "      <td>ok game.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2</td>\n",
       "      <td>A2204E1TH211HT</td>\n",
       "      <td>0700026657</td>\n",
       "      <td>Grandma KR</td>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497339</th>\n",
       "      <td>4</td>\n",
       "      <td>AVECM71LSZLC5</td>\n",
       "      <td>B01HGPUTCA</td>\n",
       "      <td>boris teplitskiy</td>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497340</th>\n",
       "      <td>3</td>\n",
       "      <td>A1RS06313BL6WN</td>\n",
       "      <td>B01HH6JEOC</td>\n",
       "      <td>Tom Stopsign</td>\n",
       "      <td>Okay stuff.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497341</th>\n",
       "      <td>3</td>\n",
       "      <td>ACIZ77IGIX2JL</td>\n",
       "      <td>B01HH6JEOC</td>\n",
       "      <td>Era</td>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497342</th>\n",
       "      <td>4</td>\n",
       "      <td>A34GG58TJ1A3SH</td>\n",
       "      <td>B01HIZF7XE</td>\n",
       "      <td>seamonkey10</td>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497343</th>\n",
       "      <td>2</td>\n",
       "      <td>A6W81WTFK940B</td>\n",
       "      <td>B01HIZGKOE</td>\n",
       "      <td>msam420</td>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497344 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              0               1           2                  3  \\\n",
       "0       overall      reviewerID        asin       reviewerName   \n",
       "1             5  A1HP7NVNPFMA4N  0700026657        Ambrosia075   \n",
       "2             4  A1JGAP0185YJI6  0700026657             travis   \n",
       "3             3  A1YJWEXHQBWK2B  0700026657  Vincent G. Mezera   \n",
       "4             2  A2204E1TH211HT  0700026657         Grandma KR   \n",
       "...         ...             ...         ...                ...   \n",
       "497339        4   AVECM71LSZLC5  B01HGPUTCA   boris teplitskiy   \n",
       "497340        3  A1RS06313BL6WN  B01HH6JEOC       Tom Stopsign   \n",
       "497341        3   ACIZ77IGIX2JL  B01HH6JEOC                Era   \n",
       "497342        4  A34GG58TJ1A3SH  B01HIZF7XE        seamonkey10   \n",
       "497343        2   A6W81WTFK940B  B01HIZGKOE            msam420   \n",
       "\n",
       "                                                        4  \n",
       "0                                              reviewText  \n",
       "1       This game is a bit hard to get the hang of, bu...  \n",
       "2       I played it a while but it was alright. The st...  \n",
       "3                                                ok game.  \n",
       "4       found the game a bit too complicated, not what...  \n",
       "...                                                   ...  \n",
       "497339                 not OEM but good replacement parts  \n",
       "497340                                        Okay stuff.  \n",
       "497341  This does add some kids room things that are v...  \n",
       "497342  I think I originally began playing Bioshock se...  \n",
       "497343  The graphics are terrible, it looks like ps2 g...  \n",
       "\n",
       "[497344 rows x 5 columns]"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df['text'] = train_df.iloc[:, 4]\n",
    "train_df['label'] = train_df.iloc[:, 0]\n",
    "train_df = train_df.drop(train_df.columns[[0,1, 2, 3, 4]], axis=1)\n",
    "#train_df = train_df.drop(train_df.rows[[0]], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>reviewText</td>\n",
       "      <td>overall</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ok game.</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497339</th>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497340</th>\n",
       "      <td>Okay stuff.</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497341</th>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497342</th>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497343</th>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497344 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     text    label\n",
       "0                                              reviewText  overall\n",
       "1       This game is a bit hard to get the hang of, bu...        5\n",
       "2       I played it a while but it was alright. The st...        4\n",
       "3                                                ok game.        3\n",
       "4       found the game a bit too complicated, not what...        2\n",
       "...                                                   ...      ...\n",
       "497339                 not OEM but good replacement parts        4\n",
       "497340                                        Okay stuff.        3\n",
       "497341  This does add some kids room things that are v...        3\n",
       "497342  I think I originally began playing Bioshock se...        4\n",
       "497343  The graphics are terrible, it looks like ps2 g...        2\n",
       "\n",
       "[497344 rows x 2 columns]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df.columns = ['text', 'label']\n",
    "train_df = train_df[['text', 'label']]\n",
    "#train_df['text'] = train_df['text'].apply(lambda x: x.replace('\\\\', ' '))\n",
    "#train_df['label'] = train_df['label'].apply(lambda x:str(x)-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>reviewText</td>\n",
       "      <td>overall</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ok game.</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497339</th>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497340</th>\n",
       "      <td>Okay stuff.</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497341</th>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497342</th>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497343</th>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497344 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     text    label\n",
       "0                                              reviewText  overall\n",
       "1       This game is a bit hard to get the hang of, bu...        5\n",
       "2       I played it a while but it was alright. The st...        4\n",
       "3                                                ok game.        3\n",
       "4       found the game a bit too complicated, not what...        2\n",
       "...                                                   ...      ...\n",
       "497339                 not OEM but good replacement parts        4\n",
       "497340                                        Okay stuff.        3\n",
       "497341  This does add some kids room things that are v...        3\n",
       "497342  I think I originally began playing Bioshock se...        4\n",
       "497343  The graphics are terrible, it looks like ps2 g...        2\n",
       "\n",
       "[497344 rows x 2 columns]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/IPython/core/interactiveshell.py:3146: DtypeWarning: Columns (0) have mixed types.Specify dtype option on import or set low_memory=False.\n",
      "  interactivity=interactivity, compiler=compiler, result=result)\n"
     ]
    }
   ],
   "source": [
    "eval_df = pd.read_csv('Downloads/Video_Games_5csv.csv', header=None)\n",
    "eval_df.sample(frac=1)\n",
    "eval_df['text'] = eval_df.iloc[:, 4]\n",
    "eval_df['label'] = eval_df.iloc[:, 0]\n",
    "eval_df = eval_df.drop(eval_df.columns[[0,1, 2, 3, 4]], axis=1)\n",
    "#eval_df.columns = ['text', 'label']\n",
    "#eval_df = eval_df[['text', 'label']]\n",
    "#eval_df['text'] = eval_df['text'].apply(lambda x: x.replace('\\\\', ' '))\n",
    "#eval_df['label'] = eval_df['label'].apply(lambda x:x-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "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>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>reviewText</td>\n",
       "      <td>overall</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ok game.</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497339</th>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497340</th>\n",
       "      <td>Okay stuff.</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497341</th>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497342</th>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497343</th>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497344 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     text    label\n",
       "0                                              reviewText  overall\n",
       "1       This game is a bit hard to get the hang of, bu...        5\n",
       "2       I played it a while but it was alright. The st...        4\n",
       "3                                                ok game.        3\n",
       "4       found the game a bit too complicated, not what...        2\n",
       "...                                                   ...      ...\n",
       "497339                 not OEM but good replacement parts        4\n",
       "497340                                        Okay stuff.        3\n",
       "497341  This does add some kids room things that are v...        3\n",
       "497342  I think I originally began playing Bioshock se...        4\n",
       "497343  The graphics are terrible, it looks like ps2 g...        2\n",
       "\n",
       "[497344 rows x 2 columns]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "eval_df = eval_df.iloc[1:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "eval_df  = eval_df.dropna(axis = 0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "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>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ok game.</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>great game, I love it and have played it since...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497339</th>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497340</th>\n",
       "      <td>Okay stuff.</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497341</th>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497342</th>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497343</th>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497339 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     text label\n",
       "1       This game is a bit hard to get the hang of, bu...     5\n",
       "2       I played it a while but it was alright. The st...     4\n",
       "3                                                ok game.     3\n",
       "4       found the game a bit too complicated, not what...     2\n",
       "5       great game, I love it and have played it since...     5\n",
       "...                                                   ...   ...\n",
       "497339                 not OEM but good replacement parts     4\n",
       "497340                                        Okay stuff.     3\n",
       "497341  This does add some kids room things that are v...     3\n",
       "497342  I think I originally began playing Bioshock se...     4\n",
       "497343  The graphics are terrible, it looks like ps2 g...     2\n",
       "\n",
       "[497339 rows x 2 columns]"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "eval_df['label'] = eval_df['label'].apply(lambda x:int(x)-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "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>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ok game.</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>great game, I love it and have played it since...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497339</th>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497340</th>\n",
       "      <td>Okay stuff.</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497341</th>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497342</th>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497343</th>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497339 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     text  label\n",
       "1       This game is a bit hard to get the hang of, bu...      4\n",
       "2       I played it a while but it was alright. The st...      3\n",
       "3                                                ok game.      2\n",
       "4       found the game a bit too complicated, not what...      1\n",
       "5       great game, I love it and have played it since...      4\n",
       "...                                                   ...    ...\n",
       "497339                 not OEM but good replacement parts      3\n",
       "497340                                        Okay stuff.      2\n",
       "497341  This does add some kids room things that are v...      2\n",
       "497342  I think I originally began playing Bioshock se...      3\n",
       "497343  The graphics are terrible, it looks like ps2 g...      1\n",
       "\n",
       "[497339 rows x 2 columns]"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df = eval_df.iloc[0:5000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "eval_dftrun = eval_df.iloc[5000:6000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "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>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>This game is a bit hard to get the hang of, bu...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>I played it a while but it was alright. The st...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ok game.</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>found the game a bit too complicated, not what...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>great game, I love it and have played it since...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497339</th>\n",
       "      <td>not OEM but good replacement parts</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497340</th>\n",
       "      <td>Okay stuff.</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497341</th>\n",
       "      <td>This does add some kids room things that are v...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497342</th>\n",
       "      <td>I think I originally began playing Bioshock se...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497343</th>\n",
       "      <td>The graphics are terrible, it looks like ps2 g...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>497339 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     text  label\n",
       "1       This game is a bit hard to get the hang of, bu...      4\n",
       "2       I played it a while but it was alright. The st...      3\n",
       "3                                                ok game.      2\n",
       "4       found the game a bit too complicated, not what...      1\n",
       "5       great game, I love it and have played it since...      4\n",
       "...                                                   ...    ...\n",
       "497339                 not OEM but good replacement parts      3\n",
       "497340                                        Okay stuff.      2\n",
       "497341  This does add some kids room things that are v...      2\n",
       "497342  I think I originally began playing Bioshock se...      3\n",
       "497343  The graphics are terrible, it looks like ps2 g...      1\n",
       "\n",
       "[497339 rows x 2 columns]"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at xlnet-base-cased were not used when initializing XLNetForSequenceClassification: ['lm_loss.weight', 'lm_loss.bias']\n",
      "- This IS expected if you are initializing XLNetForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing XLNetForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "Some weights of XLNetForSequenceClassification were not initialized from the model checkpoint at xlnet-base-cased and are newly initialized: ['sequence_summary.summary.weight', 'sequence_summary.summary.bias', 'logits_proj.weight', 'logits_proj.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "source": [
    "from simpletransformers.classification import ClassificationModel\n",
    "\n",
    "\n",
    "# Create a ClassificationModel\n",
    "model = ClassificationModel('xlnet', 'xlnet-base-cased', num_labels=5, use_cuda = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4f2335746ea240abb861d8d72a3e2757",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=5000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4f0cc6b9b0114271805b421d3f97868d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch', max=1.0, style=ProgressStyle(description_width='i…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b6baf1be2e024a0e8f25b7c5644d1c47",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Epoch 0 of 1', max=625.0, style=ProgressStyle(des…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n"
     ]
    },
    {
     "ename": "RuntimeError",
     "evalue": "[enforce fail at inline_container.cc:262] . unexpected pos 40165248 vs 40165136",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m--------------------------------------\u001b[0m",
      "\u001b[0;31mOSError\u001b[0mTraceback (most recent call last)",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/serialization.py\u001b[0m in \u001b[0;36msave\u001b[0;34m(obj, f, pickle_module, pickle_protocol, _use_new_zipfile_serialization)\u001b[0m\n\u001b[1;32m    363\u001b[0m             \u001b[0;32mwith\u001b[0m \u001b[0m_open_zipfile_writer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mopened_file\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mopened_zipfile\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 364\u001b[0;31m                 \u001b[0m_save\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mobj\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mopened_zipfile\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpickle_module\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpickle_protocol\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    365\u001b[0m                 \u001b[0;32mreturn\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/serialization.py\u001b[0m in \u001b[0;36m_save\u001b[0;34m(obj, zip_file, pickle_module, pickle_protocol)\u001b[0m\n\u001b[1;32m    476\u001b[0m             \u001b[0mnum_bytes\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mstorage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mstorage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0melement_size\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 477\u001b[0;31m             \u001b[0mzip_file\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwrite_record\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstorage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata_ptr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_bytes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    478\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mOSError\u001b[0m: [Errno 28] No space left on device",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[0;31mRuntimeError\u001b[0mTraceback (most recent call last)",
      "\u001b[0;32m<ipython-input-44-ffb9164a3ad9>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;31m# Train the model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_df\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/simpletransformers/classification/classification_model.py\u001b[0m in \u001b[0;36mtrain_model\u001b[0;34m(self, train_df, multi_label, output_dir, show_running_loss, args, eval_df, verbose, **kwargs)\u001b[0m\n\u001b[1;32m    402\u001b[0m             \u001b[0meval_df\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0meval_df\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    403\u001b[0m             \u001b[0mverbose\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mverbose\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 404\u001b[0;31m             \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    405\u001b[0m         )\n\u001b[1;32m    406\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/simpletransformers/classification/classification_model.py\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(self, train_dataloader, output_dir, multi_label, show_running_loss, eval_df, verbose, **kwargs)\u001b[0m\n\u001b[1;32m    727\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    728\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave_model_every_epoch\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 729\u001b[0;31m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutput_dir_current\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moptimizer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mscheduler\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    730\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    731\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate_during_training\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate_each_epoch\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/simpletransformers/classification/classification_model.py\u001b[0m in \u001b[0;36msave_model\u001b[0;34m(self, output_dir, optimizer, scheduler, model, results)\u001b[0m\n\u001b[1;32m   1543\u001b[0m             \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mos\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mjoin\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutput_dir\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"training_args.bin\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1544\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0moptimizer\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mscheduler\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave_optimizer_and_scheduler\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1545\u001b[0;31m                 \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstate_dict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mos\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mjoin\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutput_dir\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"optimizer.pt\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1546\u001b[0m                 \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mscheduler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstate_dict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mos\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mjoin\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutput_dir\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"scheduler.pt\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1547\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave_model_args\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutput_dir\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/serialization.py\u001b[0m in \u001b[0;36msave\u001b[0;34m(obj, f, pickle_module, pickle_protocol, _use_new_zipfile_serialization)\u001b[0m\n\u001b[1;32m    363\u001b[0m             \u001b[0;32mwith\u001b[0m \u001b[0m_open_zipfile_writer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mopened_file\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mopened_zipfile\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    364\u001b[0m                 \u001b[0m_save\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mobj\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mopened_zipfile\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpickle_module\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpickle_protocol\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 365\u001b[0;31m                 \u001b[0;32mreturn\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    366\u001b[0m         \u001b[0m_legacy_save\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mobj\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mopened_file\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpickle_module\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpickle_protocol\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    367\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/serialization.py\u001b[0m in \u001b[0;36m__exit__\u001b[0;34m(self, *args)\u001b[0m\n\u001b[1;32m    256\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    257\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__exit__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 258\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfile_like\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwrite_end_of_file\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    259\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuffer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mflush\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    260\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mRuntimeError\u001b[0m: [enforce fail at inline_container.cc:262] . unexpected pos 40165248 vs 40165136"
     ]
    }
   ],
   "source": [
    "# Train the model\n",
    "model.train_model(train_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2a43e45ce2ac4b67b802d8e00eadbd0e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "597aaf3ed81a4ca28fa73582247d7455",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import f1_score, accuracy_score\n",
    "\n",
    "\n",
    "def f1_multiclass(labels, preds):\n",
    "    return f1_score(labels, preds, average='micro')\n",
    "    \n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, f1=f1_multiclass, acc=accuracy_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'mcc': 0.4135980245877873,\n",
       " 'f1': 0.702,\n",
       " 'acc': 0.702,\n",
       " 'eval_loss': 0.812588570177555}"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "490af2ad644040d2b75056aa2513c6af",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ac9d6057840d47148ad2eb25ed4e9d32",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[10000:11000]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'mcc': 0.37617129324863885, 'acc': 0.701, 'eval_loss': 0.7840018681287766}"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5052f59db1e04a15bf8f1162e063404f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "634bc33fe9ba4a69bb0cf35b5e948a19",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.34645608362786556, 'acc': 0.627, 'eval_loss': 0.9404988695979118}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[50342:51342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ebff9dc2acc94cdcbc265aeef22b6cf9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b3697f1cb449482a89d72a8def0ea581",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.3863914074686631, 'acc': 0.612, 'eval_loss': 1.0088129776716233}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[150342:151342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4ff59715d2224afdab2da11ec8e35cac",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "078ab9ed0cce48baa3e78a7d0a4a2da2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.4407023503427808, 'acc': 0.699, 'eval_loss': 0.8224271050393581}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[450342:451342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "191b4a95a1ea4a74a521c47f14827116",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ba8716285cc54aaa991976e594241a74",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.40647336215227653, 'acc': 0.672, 'eval_loss': 0.8827728629112244}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[407342:408342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b6b90bc8173d4ac6bc1f62edfd3c4c59",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d0c1618b67e14945a65bd848693eb689",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.4194099116404517, 'acc': 0.768, 'eval_loss': 0.6840691976249218}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[357342:358342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5c93ad10483441d19390af345c0304ba",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9a6556c18214440bb6e5c835100398fa",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.4654181751498634, 'acc': 0.705, 'eval_loss': 0.7940712544620037}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[302342:303342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2e3661936b5b4443af145ac67443f77a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c24c5c326cd543188009aaf8f80b3be9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.3852043008839199, 'acc': 0.637, 'eval_loss': 0.9609800043702126}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[251342:252342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "94c6fcc443e9421c80f2f037e6a1aa5c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "486eecf92b3b4b588222f2090e33df23",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.38833618255218894, 'acc': 0.631, 'eval_loss': 0.9111251401901245}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[208342:209342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0406726fbaa14723a17705d16ffba8a7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d0a381781a1c4d0680eb0730dd361c35",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Evaluation', max=125.0, style=ProgressStyle(descr…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{'mcc': 0.33487395787781377, 'acc': 0.605, 'eval_loss': 1.036223285496235}\n"
     ]
    }
   ],
   "source": [
    "eval_dftrun = eval_df.iloc[102342:103342]\n",
    "result, model_outputs, wrong_predictions = model.eval_model(eval_dftrun, acc=accuracy_score)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# BERT pretrained model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-base-cased were not used when initializing BertForSequenceClassification: ['cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.seq_relationship.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.LayerNorm.bias']\n",
      "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-cased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "source": [
    "from simpletransformers.classification import ClassificationModel\n",
    "\n",
    "model = ClassificationModel('bert', 'bert-base-cased', num_labels=5, args={'reprocess_input_data': True, 'overwrite_output_dir': True}, use_cuda = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/simpletransformers/classification/classification_model.py:380: UserWarning: Dataframe headers not specified. Falling back to using column 0 as text and column 1 as labels.\n",
      "  \"Dataframe headers not specified. Falling back to using column 0 as text and column 1 as labels.\"\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dd018b43e3be4d31a699ac0f50e506d4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=5000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "82aadef0d14f49bd93d78e8a284278ae",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch', max=1.0, style=ProgressStyle(description_width='i…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "638ae75377444c2394c55c6e275b2930",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Running Epoch 0 of 1', max=625.0, style=ProgressStyle(des…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model.train_model(train_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.63"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "##Finding the baseline accuracies\n",
    "len([d for d in data[5000:6000] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.651"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[10000:11000] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.558"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[50342:51342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.508"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[150342:151342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.632"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[450342:451342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.628"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[407342:408342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.721"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[357342:358342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.608"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[302342:303342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.558"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[251342:252342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.532"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[208342:209342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.533"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len([d for d in data[102342:103342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "len([d for d in data[102342:103342] if d['overall'] == 5])/1000"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparing different models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "   \n",
    "iteration = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]\n",
    "XLNet = [0.702, 0.701, 0.63, 0.62, 0.699, 0.672, 0.768, 0.705, 0.637, 0.63, 0.61]\n",
    "baseline = [0.63, 0.65, 0.558, 0.5, 0.63, 0.63, 0.72, 0.6, 0.55, 0.532, 0.533]\n",
    "  \n",
    "plt.plot(iteration, XLNet, marker = 'o')\n",
    "plt.plot(iteration, baseline, color = 'red', marker = 'o')\n",
    "plt.title('XLNet (BLUE) Vs Baseline (RED) Accuracies')\n",
    "plt.xlabel('Validation Set')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "   \n",
    "iteration = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]\n",
    "XLNet = [0.702, 0.701, 0.63, 0.62, 0.699, 0.672, 0.768, 0.705, 0.637, 0.63, 0.61]\n",
    "baseline = [0.63, 0.65, 0.558, 0.5, 0.63, 0.63, 0.72, 0.6, 0.55, 0.532, 0.533]\n",
    "BoW = [0.625, 0.645, 0.61, 0.65, 0.61, 0.63, 0.614, 0.637, 0.635, 0.617, 0.63]\n",
    "  \n",
    "plt.plot(iteration, XLNet, marker = 'o')\n",
    "plt.plot(iteration, BoW, color = 'green', marker = 'o')\n",
    "plt.title('XLNet (BLUE) Vs BoW (GREEN) Accuracies')\n",
    "plt.xlabel('Validation Set')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "   \n",
    "iteration = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]\n",
    "XLNet = [0.702, 0.701, 0.63, 0.62, 0.699, 0.672, 0.768, 0.705, 0.637, 0.63, 0.61]\n",
    "baseline = [0.63, 0.65, 0.558, 0.5, 0.63, 0.63, 0.72, 0.6, 0.55, 0.532, 0.533]\n",
    "BoW = [0.625, 0.645, 0.61, 0.65, 0.61, 0.63, 0.614, 0.637, 0.635, 0.617, 0.63]\n",
    "  \n",
    "plt.plot(iteration, XLNet, marker = 'o')\n",
    "plt.plot(iteration, BoW, color = 'green', marker = 'o')\n",
    "plt.plot(iteration, baseline, color = 'red', marker = 'o')\n",
    "plt.title('XLNet (BLUE) Vs BoW (GREEN) Vs Baseline (RED) Accuracies')\n",
    "plt.xlabel('Validation Set')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5939090909090909"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import statistics\n",
    "#statistics.mean(XLNet)\n",
    "statistics.mean(baseline)"
   ]
  }
 ],
 "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}