bert_vectors_chandni.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Text categorization model using the features derived from BERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import json\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.metrics import confusion_matrix, classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "ORIGINAL_DATA_DIR = os.path.join(\"data\")\n",
    "BERT_FEATURE_DIR = \"bert_output_data\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6000, 2)\n",
      "6000\n"
     ]
    }
   ],
   "source": [
    "train_df = pd.read_csv(os.path.join(ORIGINAL_DATA_DIR, \"lang_id_train.csv\"))\n",
    "print(train_df.shape)\n",
    "\n",
    "bert_vectors_train = []\n",
    "with open(os.path.join(BERT_FEATURE_DIR, \"train.jsonlines\"), \"rt\") as infile:\n",
    "    for line in infile:\n",
    "        bert_data = json.loads(line)\n",
    "        for t in bert_data[\"features\"]:\n",
    "            # Only extract the [CLS] vector used for classification\n",
    "            if t[\"token\"] == \"[CLS]\":\n",
    "                # We only use the representation at the final layer of the network\n",
    "                bert_vectors_train.append(t[\"layers\"][0][\"values\"])\n",
    "                break\n",
    "print(len(bert_vectors_train))\n",
    "\n",
    "X_train = np.array(bert_vectors_train)\n",
    "y_train = train_df[\"native_language\"].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2000, 2)\n",
      "2000\n"
     ]
    }
   ],
   "source": [
    "eval_df = pd.read_csv(os.path.join(ORIGINAL_DATA_DIR, \"lang_id_eval.csv\"))\n",
    "print(eval_df.shape)\n",
    "\n",
    "bert_vectors_eval = []\n",
    "with open(os.path.join(BERT_FEATURE_DIR, \"eval.jsonlines\"), \"rt\") as infile:\n",
    "    for line in infile:\n",
    "        bert_data = json.loads(line)\n",
    "        for t in bert_data[\"features\"]:\n",
    "            # Only extract the [CLS] vector used for classification\n",
    "            if t[\"token\"] == \"[CLS]\":\n",
    "                # We only use the representation at the final layer of the network\n",
    "                bert_vectors_eval.append(t[\"layers\"][0][\"values\"])\n",
    "                break\n",
    "print(len(bert_vectors_eval))\n",
    "\n",
    "X_eval = np.array(bert_vectors_eval)\n",
    "y_eval = eval_df[\"native_language\"].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2000, 2)\n",
      "2000\n"
     ]
    }
   ],
   "source": [
    "test_df = pd.read_csv(os.path.join(ORIGINAL_DATA_DIR, \"lang_id_test.csv\"))\n",
    "print(test_df.shape)\n",
    "\n",
    "bert_vectors_test = []\n",
    "with open(os.path.join(BERT_FEATURE_DIR, \"test.jsonlines\"), \"rt\") as infile:\n",
    "    for line in infile:\n",
    "        bert_data = json.loads(line)\n",
    "        for t in bert_data[\"features\"]:\n",
    "            # Only extract the [CLS] vector used for classification\n",
    "            if t[\"token\"] == \"[CLS]\":\n",
    "                # We only use the representation at the final layer of the network\n",
    "                bert_vectors_test.append(t[\"layers\"][0][\"values\"])\n",
    "                break\n",
    "print(len(bert_vectors_test))\n",
    "\n",
    "X_test = np.array(bert_vectors_test)\n",
    "y_test = test_df[\"native_language\"].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Program Files (x86)\\Microsoft Visual Studio\\Shared\\Anaconda3_64\\lib\\site-packages\\sklearn\\linear_model\\logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "C:\\Program Files (x86)\\Microsoft Visual Studio\\Shared\\Anaconda3_64\\lib\\site-packages\\sklearn\\linear_model\\logistic.py:469: FutureWarning: Default multi_class will be changed to 'auto' in 0.22. Specify the multi_class option to silence this warning.\n",
      "  \"this warning.\", FutureWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuarcy:  0.7346666666666667\n"
     ]
    }
   ],
   "source": [
    "lr_model = LogisticRegression(penalty=\"l2\", C=1.0)\n",
    "lr_model.fit(X_train, y_train)\n",
    "\n",
    "print(\"Training Accuarcy: \", lr_model.score(X_train, y_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluation for each class\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "      Arabic       0.50      0.49      0.50       200\n",
      "   Cantonese       0.33      0.34      0.33       200\n",
      "    Japanese       0.46      0.49      0.48       200\n",
      "      Korean       0.48      0.46      0.47       200\n",
      "    Mandarin       0.32      0.32      0.32       200\n",
      "      Polish       0.47      0.51      0.49       200\n",
      "     Russian       0.51      0.57      0.54       200\n",
      "     Spanish       0.52      0.52      0.52       200\n",
      "        Thai       0.66      0.60      0.63       200\n",
      "  Vietnamese       0.42      0.36      0.39       200\n",
      "\n",
      "    accuracy                           0.47      2000\n",
      "   macro avg       0.47      0.47      0.47      2000\n",
      "weighted avg       0.47      0.47      0.47      2000\n",
      "\n",
      "\n",
      "**********************************************************************************************\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "**********************************************************************************************\n",
      "\n",
      "Misclassification for each class\n",
      "     Language  Misclassification\n",
      "0      Arabic              10.05\n",
      "0   Cantonese              13.60\n",
      "0    Japanese              10.85\n",
      "0      Korean              10.30\n",
      "0    Mandarin              13.60\n",
      "0      Polish              10.65\n",
      "0     Russian               9.70\n",
      "0     Spanish               9.65\n",
      "0        Thai               7.10\n",
      "0  Vietnamese              11.30\n",
      "\n",
      "**********************************************************************************************\n",
      "\n",
      "Misclassification between each pair of classes\n",
      "     Language   Predicted Misclassification\n",
      "0     Russian        Thai                 0\n",
      "0        Thai     Russian                 1\n",
      "0     Spanish   Cantonese                 2\n",
      "0        Thai      Polish                 2\n",
      "0      Polish      Korean                 3\n",
      "0   Cantonese     Spanish                 4\n",
      "0    Japanese     Spanish                 5\n",
      "0      Polish    Mandarin                 5\n",
      "0    Japanese        Thai                 5\n",
      "0     Spanish      Korean                 5\n",
      "0     Spanish        Thai                 5\n",
      "0     Russian      Korean                 5\n",
      "0     Russian    Mandarin                 6\n",
      "0    Mandarin        Thai                 6\n",
      "0     Russian  Vietnamese                 6\n",
      "0        Thai    Japanese                 7\n",
      "0      Korean      Arabic                 7\n",
      "0      Arabic      Korean                 7\n",
      "0      Korean      Polish                 8\n",
      "0      Korean  Vietnamese                 8\n",
      "0      Polish        Thai                 8\n",
      "0      Polish    Japanese                 8\n",
      "0     Russian   Cantonese                 8\n",
      "0  Vietnamese        Thai                 8\n",
      "0    Japanese      Arabic                 8\n",
      "0      Polish  Vietnamese                 9\n",
      "0    Japanese  Vietnamese                 9\n",
      "0   Cantonese        Thai                 9\n",
      "0      Arabic    Japanese                 9\n",
      "0        Thai    Mandarin                 9\n",
      "0      Korean     Spanish                 9\n",
      "0      Arabic        Thai                 9\n",
      "0    Mandarin      Polish                 9\n",
      "0   Cantonese     Russian                10\n",
      "0        Thai   Cantonese                10\n",
      "0     Russian      Arabic                10\n",
      "0    Japanese     Russian                10\n",
      "0      Arabic     Russian                10\n",
      "0    Mandarin     Russian                10\n",
      "0     Spanish    Japanese                11\n",
      "0        Thai  Vietnamese                11\n",
      "0      Polish     Spanish                11\n",
      "0     Spanish  Vietnamese                11\n",
      "0  Vietnamese    Japanese                11\n",
      "0      Polish      Arabic                11\n",
      "0      Arabic   Cantonese                11\n",
      "0  Vietnamese     Spanish                11\n",
      "0   Cantonese      Arabic                11\n",
      "0    Japanese    Mandarin                11\n",
      "0   Cantonese      Korean                12\n",
      "0   Cantonese      Polish                12\n",
      "0      Korean        Thai                12\n",
      "0      Korean     Russian                12\n",
      "0  Vietnamese      Arabic                12\n",
      "0      Arabic    Mandarin                12\n",
      "0  Vietnamese     Russian                12\n",
      "0    Mandarin      Arabic                12\n",
      "0        Thai      Arabic                12\n",
      "0      Korean    Mandarin                13\n",
      "0    Japanese   Cantonese                13\n",
      "0        Thai     Spanish                13\n",
      "0      Arabic  Vietnamese                13\n",
      "0  Vietnamese      Korean                13\n",
      "0  Vietnamese      Polish                13\n",
      "0      Polish   Cantonese                14\n",
      "0     Spanish     Russian                14\n",
      "0   Cantonese    Japanese                14\n",
      "0      Arabic      Polish                14\n",
      "0     Spanish    Mandarin                14\n",
      "0   Cantonese  Vietnamese                14\n",
      "0    Mandarin     Spanish                14\n",
      "0    Mandarin      Korean                14\n",
      "0     Russian    Japanese                14\n",
      "0     Russian     Spanish                14\n",
      "0        Thai      Korean                15\n",
      "0      Arabic     Spanish                16\n",
      "0     Spanish      Polish                17\n",
      "0    Mandarin  Vietnamese                17\n",
      "0    Japanese      Polish                17\n",
      "0     Spanish      Arabic                17\n",
      "0      Korean   Cantonese                17\n",
      "0    Mandarin    Japanese                19\n",
      "0  Vietnamese    Mandarin                19\n",
      "0     Russian      Polish                23\n",
      "0    Japanese      Korean                23\n",
      "0      Korean    Japanese                23\n",
      "0  Vietnamese   Cantonese                29\n",
      "0      Polish     Russian                29\n",
      "0    Mandarin   Cantonese                36\n",
      "0   Cantonese    Mandarin                46\n",
      "\n",
      "**********************************************************************************************\n",
      "\n",
      "Summary\n",
      "Total records: 2000\n",
      "Incorrect predictions: 1068\n",
      "Correct predictions: 932\n"
     ]
    }
   ],
   "source": [
    "# Adding predicted value in the dataframe\n",
    "test_df['predicted1'] = lr_model.predict(X_test)\n",
    "# Class list\n",
    "list_of_languages = sorted(test_df['native_language'].unique())\n",
    "# Precision, recall, f-measure and support for each class\n",
    "print(\"Evaluation for each class\")\n",
    "print(classification_report(y_test,test_df['predicted1'].values,target_names=list_of_languages))\n",
    "print()\n",
    "print(\"**********************************************************************************************\")\n",
    "print()\n",
    "\n",
    "\n",
    "# Confusion matrix\n",
    "matrix = confusion_matrix(test_df['native_language'], test_df['predicted1'])\n",
    "plt.figure(figsize = (10,5))\n",
    "ax = sns.heatmap(matrix, annot=True, xticklabels=list_of_languages, yticklabels=list_of_languages)\n",
    "plt.show()\n",
    "print(\"**********************************************************************************************\")\n",
    "print()\n",
    "\n",
    "\n",
    "# Calculate misclassification\n",
    "test_predicted = test_df.groupby('predicted1').count()['native_language']\n",
    "test_misclassifications = []\n",
    "for i in range(len(list_of_languages)):\n",
    "    misclassification = ((200 - matrix[i][i] + (test_predicted[i] - matrix[i][i])) / 2000) * 100\n",
    "    test_misclassifications.append(misclassification)\n",
    "\n",
    "# Misclassification for each class into one dataframe\n",
    "evaluation_by_class = pd.DataFrame(columns=['Language', 'Misclassification'])\n",
    "for i in range(len(list_of_languages)):\n",
    "    evaluation_by_class = evaluation_by_class.append(pd.DataFrame([[list_of_languages[i], test_misclassifications[i]]], \n",
    "                                        columns=['Language', 'Misclassification']))\n",
    "print(\"Misclassification for each class\")\n",
    "print(evaluation_by_class.to_string())\n",
    "print()\n",
    "print(\"**********************************************************************************************\")\n",
    "print()\n",
    "\n",
    "\n",
    "# Evaluate misclassification between all classes\n",
    "evaluation_between_classes = pd.DataFrame(columns=['Language', 'Predicted', 'Misclassification'])\n",
    "for i in list_of_languages:\n",
    "    for j in list_of_languages:\n",
    "        if(i != j):\n",
    "            evaluation_between_classes = evaluation_between_classes.append(pd.DataFrame([[i, j, \n",
    "                                                              matrix[list_of_languages.index(i)][list_of_languages.index(j)]]], \n",
    "                                                              columns=['Language', 'Predicted', 'Misclassification']))\n",
    "print(\"Misclassification between each pair of classes\")\n",
    "print(evaluation_between_classes.sort_values(by=['Misclassification']).to_string())\n",
    "print()\n",
    "print(\"**********************************************************************************************\")\n",
    "print()\n",
    "\n",
    "\n",
    "print(\"Summary\")\n",
    "print(\"Total records:\", test_df.shape[0])\n",
    "print(\"Incorrect predictions:\", evaluation_between_classes['Misclassification'].sum())\n",
    "print(\"Correct predictions:\", (test_df.shape[0] - evaluation_between_classes['Misclassification'].sum()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#  Neural Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuarcy:  0.6841666666666667\n"
     ]
    }
   ],
   "source": [
    "nn_model = MLPClassifier(solver='lbfgs')\n",
    "nn_model.fit(X_train, y_train) \n",
    "\n",
    "print(\"Training Accuarcy: \", nn_model.score(X_train, y_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluation for each class\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "      Arabic       0.51      0.43      0.47       200\n",
      "   Cantonese       0.33      0.34      0.34       200\n",
      "    Japanese       0.45      0.46      0.45       200\n",
      "      Korean       0.47      0.45      0.46       200\n",
      "    Mandarin       0.32      0.34      0.33       200\n",
      "      Polish       0.44      0.47      0.46       200\n",
      "     Russian       0.50      0.55      0.52       200\n",
      "     Spanish       0.55      0.52      0.53       200\n",
      "        Thai       0.58      0.59      0.59       200\n",
      "  Vietnamese       0.39      0.38      0.38       200\n",
      "\n",
      "    accuracy                           0.45      2000\n",
      "   macro avg       0.45      0.45      0.45      2000\n",
      "weighted avg       0.45      0.45      0.45      2000\n",
      "\n",
      "\n",
      "**********************************************************************************************\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "**********************************************************************************************\n",
      "\n",
      "Misclassification for each class\n",
      "     Language  Misclassification\n",
      "0      Arabic               9.80\n",
      "0   Cantonese              13.60\n",
      "0    Japanese              11.05\n",
      "0      Korean              10.50\n",
      "0    Mandarin              13.85\n",
      "0      Polish              11.20\n",
      "0     Russian              10.05\n",
      "0     Spanish               9.05\n",
      "0        Thai               8.35\n",
      "0  Vietnamese              12.05\n",
      "\n",
      "**********************************************************************************************\n",
      "\n",
      "Misclassification between each pair of classes\n",
      "     Language   Predicted Misclassification\n",
      "0     Russian        Thai                 1\n",
      "0        Thai     Russian                 2\n",
      "0   Cantonese     Spanish                 3\n",
      "0     Russian    Mandarin                 4\n",
      "0     Russian      Korean                 5\n",
      "0      Korean     Spanish                 5\n",
      "0      Polish      Korean                 5\n",
      "0        Thai      Arabic                 6\n",
      "0      Polish    Mandarin                 6\n",
      "0     Spanish        Thai                 6\n",
      "0    Japanese     Spanish                 7\n",
      "0        Thai     Spanish                 7\n",
      "0      Korean      Polish                 7\n",
      "0        Thai      Polish                 7\n",
      "0    Japanese      Arabic                 7\n",
      "0      Korean      Arabic                 7\n",
      "0    Japanese        Thai                 7\n",
      "0     Russian   Cantonese                 7\n",
      "0     Spanish   Cantonese                 7\n",
      "0      Arabic    Japanese                 7\n",
      "0    Mandarin      Arabic                 8\n",
      "0      Polish        Thai                 8\n",
      "0      Polish  Vietnamese                 8\n",
      "0     Spanish      Korean                 8\n",
      "0     Spanish     Russian                 8\n",
      "0        Thai    Mandarin                 8\n",
      "0    Mandarin        Thai                 8\n",
      "0      Korean  Vietnamese                 8\n",
      "0   Cantonese      Polish                 8\n",
      "0      Arabic        Thai                 8\n",
      "0   Cantonese      Arabic                 8\n",
      "0   Cantonese     Russian                 8\n",
      "0  Vietnamese      Korean                 9\n",
      "0     Russian     Spanish                 9\n",
      "0  Vietnamese     Spanish                 9\n",
      "0     Spanish    Japanese                 9\n",
      "0      Arabic     Russian                 9\n",
      "0     Russian  Vietnamese                10\n",
      "0    Mandarin      Polish                10\n",
      "0    Mandarin      Korean                10\n",
      "0      Polish     Spanish                10\n",
      "0      Arabic      Korean                10\n",
      "0        Thai    Japanese                10\n",
      "0    Mandarin     Russian                11\n",
      "0    Mandarin     Spanish                11\n",
      "0  Vietnamese    Japanese                11\n",
      "0     Russian      Arabic                11\n",
      "0      Polish      Arabic                11\n",
      "0    Japanese      Polish                11\n",
      "0      Polish    Japanese                11\n",
      "0    Japanese  Vietnamese                11\n",
      "0      Arabic   Cantonese                12\n",
      "0     Spanish      Arabic                12\n",
      "0   Cantonese  Vietnamese                12\n",
      "0      Korean     Russian                12\n",
      "0      Arabic    Mandarin                12\n",
      "0        Thai      Korean                12\n",
      "0  Vietnamese      Polish                12\n",
      "0        Thai   Cantonese                13\n",
      "0  Vietnamese      Arabic                13\n",
      "0   Cantonese    Japanese                13\n",
      "0    Japanese   Cantonese                13\n",
      "0    Japanese     Russian                13\n",
      "0      Arabic  Vietnamese                14\n",
      "0      Polish   Cantonese                14\n",
      "0  Vietnamese     Russian                14\n",
      "0    Japanese    Mandarin                14\n",
      "0     Russian    Japanese                14\n",
      "0   Cantonese        Thai                15\n",
      "0     Spanish    Mandarin                15\n",
      "0     Spanish      Polish                15\n",
      "0   Cantonese      Korean                15\n",
      "0  Vietnamese        Thai                16\n",
      "0      Korean        Thai                16\n",
      "0      Korean    Mandarin                16\n",
      "0        Thai  Vietnamese                17\n",
      "0     Spanish  Vietnamese                17\n",
      "0      Arabic      Polish                18\n",
      "0      Korean    Japanese                18\n",
      "0    Mandarin  Vietnamese                19\n",
      "0    Mandarin    Japanese                19\n",
      "0  Vietnamese   Cantonese                19\n",
      "0      Korean   Cantonese                21\n",
      "0  Vietnamese    Mandarin                22\n",
      "0      Arabic     Spanish                23\n",
      "0    Japanese      Korean                26\n",
      "0     Russian      Polish                30\n",
      "0      Polish     Russian                33\n",
      "0    Mandarin   Cantonese                35\n",
      "0   Cantonese    Mandarin                49\n",
      "\n",
      "**********************************************************************************************\n",
      "\n",
      "Summary\n",
      "Total records: 2000\n",
      "Incorrect predictions: 1095\n",
      "Correct predictions: 905\n"
     ]
    }
   ],
   "source": [
    "# Adding predicted value in the dataframe\n",
    "test_df['predicted2'] = nn_model.predict(X_test)\n",
    "# Class list\n",
    "list_of_languages = sorted(test_df['native_language'].unique())\n",
    "# Precision, recall, f-measure and support for each class\n",
    "print(\"Evaluation for each class\")\n",
    "print(classification_report(y_test,test_df['predicted2'].values,target_names=list_of_languages))\n",
    "print()\n",
    "print(\"**********************************************************************************************\")\n",
    "print()\n",
    "\n",
    "\n",
    "# Confusion matrix\n",
    "matrix = confusion_matrix(test_df['native_language'], test_df['predicted2'])\n",
    "plt.figure(figsize = (10,5))\n",
    "ax = sns.heatmap(matrix, annot=True, xticklabels=list_of_languages, yticklabels=list_of_languages)\n",
    "plt.show()\n",
    "print(\"**********************************************************************************************\")\n",
    "print()\n",
    "\n",
    "\n",
    "# Calculate misclassification\n",
    "test_predicted = test_df.groupby('predicted2').count()['native_language']\n",
    "test_misclassifications = []\n",
    "for i in range(len(list_of_languages)):\n",
    "    misclassification = ((200 - matrix[i][i] + (test_predicted[i] - matrix[i][i])) / 2000) * 100\n",
    "    test_misclassifications.append(misclassification)\n",
    "\n",
    "# Misclassification for each class into one dataframe\n",
    "evaluation_by_class = pd.DataFrame(columns=['Language', 'Misclassification'])\n",
    "for i in range(len(list_of_languages)):\n",
    "    evaluation_by_class = evaluation_by_class.append(pd.DataFrame([[list_of_languages[i], test_misclassifications[i]]], \n",
    "                                        columns=['Language', 'Misclassification']))\n",
    "print(\"Misclassification for each class\")\n",
    "print(evaluation_by_class.to_string())\n",
    "print()\n",
    "print(\"**********************************************************************************************\")\n",
    "print()\n",
    "\n",
    "\n",
    "# Evaluate misclassification between all classes\n",
    "evaluation_between_classes = pd.DataFrame(columns=['Language', 'Predicted', 'Misclassification'])\n",
    "for i in list_of_languages:\n",
    "    for j in list_of_languages:\n",
    "        if(i != j):\n",
    "            evaluation_between_classes = evaluation_between_classes.append(pd.DataFrame([[i, j, \n",
    "                                                              matrix[list_of_languages.index(i)][list_of_languages.index(j)]]], \n",
    "                                                              columns=['Language', 'Predicted', 'Misclassification']))\n",
    "print(\"Misclassification between each pair of classes\")\n",
    "print(evaluation_between_classes.sort_values(by=['Misclassification']).to_string())\n",
    "print()\n",
    "print(\"**********************************************************************************************\")\n",
    "print()\n",
    "\n",
    "\n",
    "print(\"Summary\")\n",
    "print(\"Total records:\", test_df.shape[0])\n",
    "print(\"Incorrect predictions:\", evaluation_between_classes['Misclassification'].sum())\n",
    "print(\"Correct predictions:\", (test_df.shape[0] - evaluation_between_classes['Misclassification'].sum()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}