5_audio_classification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import xgboost as xgb\n",
    "import pickle\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.svm import SVC, LinearSVC\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.metrics import confusion_matrix, f1_score, accuracy_score, precision_score, recall_score\n",
    "from sklearn.preprocessing import MinMaxScaler, OneHotEncoder\n",
    "from sklearn.feature_selection import SelectFromModel\n",
    "from sklearn.utils.class_weight import compute_class_weight\n",
    "import itertools\n",
    "from IPython.display import display\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7837, 10)\n"
     ]
    }
   ],
   "source": [
    "x_train = pd.read_csv('data/s2e/audio_train.csv')\n",
    "x_test = pd.read_csv('data/s2e/audio_test.csv')\n",
    "\n",
    "print(x_train.shape)\n",
    "y_train = x_train['label']\n",
    "y_test = x_test['label']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7837, 10) (1960, 10)\n",
      "{0: 879, 1: 1336, 2: 2350, 3: 993, 4: 936, 5: 1343}\n"
     ]
    }
   ],
   "source": [
    "print(x_train.shape, x_test.shape)\n",
    "cl_weight = dict(pd.Series(x_train['label']).value_counts(normalize=True))\n",
    "print(dict(pd.Series(x_train['label']).value_counts()))\n",
    "\n",
    "del x_train['label']\n",
    "del x_test['label']\n",
    "del x_train['wav_file']\n",
    "del x_test['wav_file']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "emotion_dict = {'ang': 0,\n",
    "                'hap': 1,\n",
    "                'sad': 2,\n",
    "                'fea': 3,\n",
    "                'sur': 4,\n",
    "                'neu': 5}\n",
    "\n",
    "emo_keys = list(['ang', 'hap', 'sad', 'fea', 'sur', 'neu'])\n",
    "\n",
    "def plot_confusion_matrix(cm, classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    \"\"\"\n",
    "    # plt.figure(figsize=(8,8))\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=45)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "        print(\"Normalized confusion matrix\")\n",
    "    else:\n",
    "        print('Confusion matrix, without normalization')\n",
    "\n",
    "    print(cm)\n",
    "\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, cm[i, j],\n",
    "                 horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')\n",
    "    \n",
    "def one_hot_encoder(true_labels, num_records, num_classes):\n",
    "    temp = np.array(true_labels[:num_records])\n",
    "    true_labels = np.zeros((num_records, num_classes))\n",
    "    true_labels[np.arange(num_records), temp] = 1\n",
    "    return true_labels\n",
    "\n",
    "def display_results(y_test, pred_probs, cm=True):\n",
    "    pred = np.argmax(pred_probs, axis=-1)\n",
    "    one_hot_true = one_hot_encoder(y_test, len(pred), len(emotion_dict))\n",
    "    print('Test Set Accuracy =  {0:.3f}'.format(accuracy_score(y_test, pred)))\n",
    "    print('Test Set F-score =  {0:.3f}'.format(f1_score(y_test, pred, average='macro')))\n",
    "    print('Test Set Precision =  {0:.3f}'.format(precision_score(y_test, pred, average='macro')))\n",
    "    print('Test Set Recall =  {0:.3f}'.format(recall_score(y_test, pred, average='macro')))\n",
    "    if cm:\n",
    "        plot_confusion_matrix(confusion_matrix(y_test, pred), classes=emo_keys)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy =  0.571\n",
      "Test Set F-score =  0.566\n",
      "Test Set Precision =  0.585\n",
      "Test Set Recall =  0.576\n",
      "Confusion matrix, without normalization\n",
      "[[ 76  41  86   2   5  14]\n",
      " [ 36  46 165   2   2  49]\n",
      " [ 25  57 425   0  14  62]\n",
      " [  0   0   0 247   0   0]\n",
      " [  0   0   0   0 241   0]\n",
      " [ 16  26 227   0  11  85]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "rf_classifier = RandomForestClassifier(n_estimators=1200, min_samples_split=25)\n",
    "rf_classifier.fit(x_train, y_train)\n",
    "\n",
    "# Predict\n",
    "pred_probs = rf_classifier.predict_proba(x_test)\n",
    "\n",
    "# Results\n",
    "display_results(y_test, pred_probs)\n",
    "\n",
    "with open('pred_probas/rf_classifier.pkl', 'wb') as f:\n",
    "    pickle.dump(pred_probs, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy =  0.565\n",
      "Test Set F-score =  0.564\n",
      "Test Set Precision =  0.578\n",
      "Test Set Recall =  0.572\n",
      "Confusion matrix, without normalization\n",
      "[[ 67  53  75   3   5  21]\n",
      " [ 36  63 149   0   3  49]\n",
      " [ 31  56 407   1  11  77]\n",
      " [  0   0   0 247   0   0]\n",
      " [  0   0   0   0 241   0]\n",
      " [ 21  20 233   2   7  82]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "xgb_classifier = xgb.XGBClassifier(max_depth=7, learning_rate=0.008, objective='multi:softprob', \n",
    "                                   n_estimators=1200, sub_sample=0.8, num_class=len(emotion_dict),\n",
    "                                   booster='gbtree', n_jobs=4)\n",
    "xgb_classifier.fit(x_train, y_train)\n",
    "\n",
    "# Predict\n",
    "pred_probs = xgb_classifier.predict_proba(x_test)\n",
    "\n",
    "# Results\n",
    "display_results(y_test, pred_probs)\n",
    "\n",
    "with open('pred_probas/xgb_classifier.pkl', 'wb') as f:\n",
    "    pickle.dump(pred_probs, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy =  0.404\n",
      "Test Set F-score =  0.352\n",
      "Test Set Precision =  0.445\n",
      "Test Set Recall =  0.355\n",
      "Confusion matrix, without normalization\n",
      "[[ 64  31  97  17  13   2]\n",
      " [ 20  52 188  13   7  20]\n",
      " [ 16  38 443  49  17  20]\n",
      " [  0   0  98 138   7   4]\n",
      " [  7  11 118  39  64   2]\n",
      " [  7  20 263  29  15  31]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mlp_classifier = MLPClassifier(hidden_layer_sizes=(650, ), activation='relu', solver='adam', alpha=0.0001,\n",
    "                               batch_size='auto', learning_rate='adaptive', learning_rate_init=0.01,\n",
    "                               power_t=0.5, max_iter=1000, shuffle=True, random_state=None, tol=0.0001,\n",
    "                               verbose=False, warm_start=True, momentum=0.8, nesterovs_momentum=True,\n",
    "                               early_stopping=False, validation_fraction=0.1, beta_1=0.9, beta_2=0.999,\n",
    "                               epsilon=1e-08)\n",
    "\n",
    "mlp_classifier.fit(x_train, y_train)\n",
    "\n",
    "# Predict\n",
    "pred_probs = mlp_classifier.predict_proba(x_test)\n",
    "\n",
    "# Results\n",
    "display_results(y_test, pred_probs)\n",
    "\n",
    "with open('pred_probas/mlp_classifier.pkl', 'wb') as f:\n",
    "    pickle.dump(pred_probs, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy =  0.404\n",
      "Test Set F-score =  0.352\n",
      "Test Set Precision =  0.445\n",
      "Test Set Recall =  0.355\n",
      "Confusion matrix, without normalization\n",
      "[[ 64  31  97  17  13   2]\n",
      " [ 20  52 188  13   7  20]\n",
      " [ 16  38 443  49  17  20]\n",
      " [  0   0  98 138   7   4]\n",
      " [  7  11 118  39  64   2]\n",
      " [  7  20 263  29  15  31]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "svc_classifier = LinearSVC()\n",
    "\n",
    "svc_classifier.fit(x_train, y_train)\n",
    "\n",
    "# Predict\n",
    "pred = svc_classifier.predict(x_test)\n",
    "\n",
    "# Results\n",
    "display_results(y_test, pred_probs)\n",
    "\n",
    "with open('pred_probas/svc_classifier_model.pkl', 'wb') as f:\n",
    "    pickle.dump(svc_classifier, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy =  0.301\n",
      "Test Set F-score =  0.089\n",
      "Test Set Precision =  0.144\n",
      "Test Set Recall =  0.173\n",
      "Confusion matrix, without normalization\n",
      "[[  9   3 212   0   0   0]\n",
      " [  3   0 297   0   0   0]\n",
      " [  2   0 581   0   0   0]\n",
      " [  0   0 247   0   0   0]\n",
      " [  2   0 239   0   0   0]\n",
      " [  0   0 365   0   0   0]]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gsahu/dev/lib/python3.5/site-packages/sklearn/metrics/classification.py:1143: UndefinedMetricWarning: F-score is ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n",
      "/home/gsahu/dev/lib/python3.5/site-packages/sklearn/metrics/classification.py:1143: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mnb_classifier = MultinomialNB()\n",
    "\n",
    "mnb_classifier.fit(x_train, y_train)\n",
    "\n",
    "# Predict\n",
    "pred_probs = mnb_classifier.predict_proba(x_test)\n",
    "\n",
    "# Results\n",
    "display_results(y_test, pred_probs)\n",
    "\n",
    "with open('pred_probas/mnb_classifier.pkl', 'wb') as f:\n",
    "    pickle.dump(pred_probs, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy =  0.323\n",
      "Test Set F-score =  0.151\n",
      "Test Set Precision =  0.190\n",
      "Test Set Recall =  0.212\n",
      "Confusion matrix, without normalization\n",
      "[[ 58  15 151   0   0   0]\n",
      " [ 17  16 267   0   0   0]\n",
      " [ 12  12 559   0   0   0]\n",
      " [  0   0 247   0   0   0]\n",
      " [ 23   1 217   0   0   0]\n",
      " [  6   5 354   0   0   0]]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/gsahu/dev/lib/python3.5/site-packages/sklearn/metrics/classification.py:1143: UndefinedMetricWarning: F-score is ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n",
      "/home/gsahu/dev/lib/python3.5/site-packages/sklearn/metrics/classification.py:1143: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr_classifier = LogisticRegression(solver='lbfgs', multi_class='multinomial', max_iter=1000)\n",
    "\n",
    "lr_classifier.fit(x_train, y_train)\n",
    "\n",
    "# Predict\n",
    "pred_probs = lr_classifier.predict_proba(x_test)\n",
    "\n",
    "# Results\n",
    "display_results(y_test, pred_probs)\n",
    "\n",
    "with open('pred_probas/lr_classifier.pkl', 'wb') as f:\n",
    "    pickle.dump(pred_probs, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax = xgb.plot_importance(xgb_classifier, max_num_features=10, height=0.5, show_values=False)\n",
    "fig = ax.figure\n",
    "fig.set_size_inches(8, 8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.10619742, 0.11682408, 0.14686272, 0.15600932, 0.11254673,\n",
       "       0.09727044, 0.14459886, 0.11969043], dtype=float32)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "contribution_scores = xgb_classifier.feature_importances_\n",
    "contribution_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stats for top 3 features:\n",
      "Test Set Accuracy =  0.537\n",
      "Test Set F-score =  0.534\n",
      "Test Set Precision =  0.549\n",
      "Test Set Recall =  0.547\n",
      "Stats for top 5 features:\n",
      "Test Set Accuracy =  0.554\n",
      "Test Set F-score =  0.548\n",
      "Test Set Precision =  0.559\n",
      "Test Set Recall =  0.561\n",
      "Stats for top 6 features:\n",
      "Test Set Accuracy =  0.559\n",
      "Test Set F-score =  0.559\n",
      "Test Set Precision =  0.570\n",
      "Test Set Recall =  0.568\n",
      "Stats for top 7 features:\n",
      "Test Set Accuracy =  0.563\n",
      "Test Set F-score =  0.560\n",
      "Test Set Precision =  0.572\n",
      "Test Set Recall =  0.570\n",
      "Stats for top 8 features:\n",
      "Test Set Accuracy =  0.565\n",
      "Test Set F-score =  0.564\n",
      "Test Set Precision =  0.578\n",
      "Test Set Recall =  0.572\n"
     ]
    }
   ],
   "source": [
    "top_n = [3, 5, 6, 7, 8]  # number of features\n",
    "for n in top_n:\n",
    "    threshold = np.argsort(contribution_scores)[::-1][:n][-1]\n",
    "    print('Stats for top {} features:'.format(n))\n",
    "    # Select features using threshold\n",
    "    selection = SelectFromModel(xgb_classifier, threshold=contribution_scores[threshold], prefit=True)\n",
    "    select_x_train = selection.transform(x_train)\n",
    "    select_x_test = selection.transform(x_test)\n",
    "    \n",
    "    # Train\n",
    "    select_xgb_classifier = xgb.XGBClassifier(max_depth=7, learning_rate=0.008, objective='multi:softprob', \n",
    "                                              n_estimators=1200, sub_sample = 0.8, num_class = len(emotion_dict),\n",
    "                                              booster='gbtree', n_jobs=4)\n",
    "    select_xgb_classifier.fit(select_x_train, y_train)\n",
    "\n",
    "    # Predict\n",
    "    pred_probs = select_xgb_classifier.predict_proba(select_x_test)\n",
    "\n",
    "    # Results\n",
    "    display_results(y_test, pred_probs, cm = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy =  0.564\n",
      "Test Set F-score =  0.550\n",
      "Test Set Precision =  0.580\n",
      "Test Set Recall =  0.562\n",
      "Confusion matrix, without normalization\n",
      "[[ 67  40  93   6   6  12]\n",
      " [ 34  48 175   1   3  39]\n",
      " [ 24  48 444   5  10  52]\n",
      " [  0   0   0 247   0   0]\n",
      " [  0   0   4   0 237   0]\n",
      " [ 12  20 263   0   8  62]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Load predicted probabilities\n",
    "# Load predicted probabilities\n",
    "with open('pred_probas/rf_classifier.pkl', 'rb') as f:\n",
    "    rf_pred_probs = pickle.load(f)\n",
    "\n",
    "with open('pred_probas/xgb_classifier.pkl', 'rb') as f:\n",
    "    xgb_pred_probs = pickle.load(f)\n",
    "    \n",
    "with open('pred_probas/svc_classifier_model.pkl', 'rb') as f:\n",
    "    svc_preds = pickle.load(f)\n",
    "    \n",
    "with open('pred_probas/mnb_classifier.pkl', 'rb') as f:\n",
    "    mnb_pred_probs = pickle.load(f)\n",
    "    \n",
    "with open('pred_probas/mlp_classifier.pkl', 'rb') as f:\n",
    "    mlp_pred_probs = pickle.load(f)\n",
    "    \n",
    "with open('pred_probas/lr_classifier.pkl', 'rb') as f:\n",
    "    lr_pred_probs = pickle.load(f)\n",
    "\n",
    "with open('pred_probas/lstm_classifier.pkl', 'rb') as f:\n",
    "    lstm_pred_probs = pickle.load(f)\n",
    "\n",
    "# Average of the predicted probabilites\n",
    "ensemble_pred_probs = (xgb_pred_probs +\n",
    "                       mlp_pred_probs +\n",
    "                       rf_pred_probs)/3.0\n",
    "# Show metrics\n",
    "display_results(y_test, ensemble_pred_probs)"
   ]
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}