decision_tree.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import recall_score\n",
    "from sklearn.metrics import precision_score\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.metrics import classification_report\n",
    "import json"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load the Training, Testing and Evaluation Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the data\n",
    "X_train = pd.read_pickle('model/X_train.pkl')\n",
    "y_train = pd.read_pickle('model/y_train.pkl')\n",
    "X_test = pd.read_pickle('model/X_test.pkl')\n",
    "y_test = pd.read_pickle('model/y_test.pkl')\n",
    "X_recommend = pd.read_pickle('model/X_recommend.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>popularity</th>\n",
       "      <th>explicit</th>\n",
       "      <th>duration_ms</th>\n",
       "      <th>danceability</th>\n",
       "      <th>energy</th>\n",
       "      <th>key</th>\n",
       "      <th>loudness</th>\n",
       "      <th>mode</th>\n",
       "      <th>speechiness</th>\n",
       "      <th>acousticness</th>\n",
       "      <th>...</th>\n",
       "      <th>urban contemporary</th>\n",
       "      <th>vallenato</th>\n",
       "      <th>vapor trap</th>\n",
       "      <th>vapor twitch</th>\n",
       "      <th>video game music</th>\n",
       "      <th>viral pop</th>\n",
       "      <th>viral rap</th>\n",
       "      <th>weirdcore</th>\n",
       "      <th>west coast rap</th>\n",
       "      <th>yacht rock</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1213</th>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>220653</td>\n",
       "      <td>0.780</td>\n",
       "      <td>0.856</td>\n",
       "      <td>10</td>\n",
       "      <td>-5.316</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0405</td>\n",
       "      <td>0.11600</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71</th>\n",
       "      <td>84</td>\n",
       "      <td>False</td>\n",
       "      <td>173333</td>\n",
       "      <td>0.726</td>\n",
       "      <td>0.431</td>\n",
       "      <td>8</td>\n",
       "      <td>-8.765</td>\n",
       "      <td>0</td>\n",
       "      <td>0.1350</td>\n",
       "      <td>0.73100</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1058</th>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>182506</td>\n",
       "      <td>0.791</td>\n",
       "      <td>0.907</td>\n",
       "      <td>0</td>\n",
       "      <td>-2.223</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0884</td>\n",
       "      <td>0.01320</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>756</th>\n",
       "      <td>77</td>\n",
       "      <td>True</td>\n",
       "      <td>205158</td>\n",
       "      <td>0.533</td>\n",
       "      <td>0.858</td>\n",
       "      <td>2</td>\n",
       "      <td>-4.212</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0766</td>\n",
       "      <td>0.00021</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>317</th>\n",
       "      <td>62</td>\n",
       "      <td>False</td>\n",
       "      <td>109735</td>\n",
       "      <td>0.670</td>\n",
       "      <td>0.609</td>\n",
       "      <td>7</td>\n",
       "      <td>-10.088</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0333</td>\n",
       "      <td>0.74900</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 316 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      popularity  explicit  duration_ms  danceability  energy  key  loudness  \\\n",
       "1213           4     False       220653         0.780   0.856   10    -5.316   \n",
       "71            84     False       173333         0.726   0.431    8    -8.765   \n",
       "1058           0     False       182506         0.791   0.907    0    -2.223   \n",
       "756           77      True       205158         0.533   0.858    2    -4.212   \n",
       "317           62     False       109735         0.670   0.609    7   -10.088   \n",
       "\n",
       "      mode  speechiness  acousticness  ...  urban contemporary  vallenato  \\\n",
       "1213     0       0.0405       0.11600  ...                   0          0   \n",
       "71       0       0.1350       0.73100  ...                   0          0   \n",
       "1058     1       0.0884       0.01320  ...                   0          0   \n",
       "756      1       0.0766       0.00021  ...                   0          0   \n",
       "317      1       0.0333       0.74900  ...                   0          0   \n",
       "\n",
       "      vapor trap  vapor twitch  video game music  viral pop  viral rap  \\\n",
       "1213           0             0                 0          0          0   \n",
       "71             0             0                 0          0          0   \n",
       "1058           0             0                 0          0          0   \n",
       "756            0             0                 0          0          0   \n",
       "317            0             0                 0          0          0   \n",
       "\n",
       "      weirdcore  west coast rap  yacht rock  \n",
       "1213          0               0           0  \n",
       "71            0               0           0  \n",
       "1058          0               0           0  \n",
       "756           0               0           0  \n",
       "317           0               0           0  \n",
       "\n",
       "[5 rows x 316 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since Decision Tree is a Tree-based algorithm, it is fairly insensitive to the scale of the features, thus we wont scale the features for this model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Hyperparameter Tuning\n",
    "\n",
    "We will tune for:\n",
    "- Function to measure the quality of a split (criterion)\n",
    "- The strategy used to choose the split at each node (splitter)\n",
    "- The maximum depth of the tree (max_depth)\n",
    "- The minimum number of samples required to be at a leaf node (min_samples_leaf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'criterion': ['gini', 'entropy'], 'splitter': ['best', 'random'], 'max_depth': [2, 3, 5, 10, 20], 'min_samples_leaf': [5, 10, 20, 50, 100]}\n"
     ]
    }
   ],
   "source": [
    "# Check accuracy for each value for each hyperparameter in its range\n",
    "# We will use GridSearchCV to automate this process\n",
    "\n",
    "criterion = [\"gini\", \"entropy\"]\n",
    "splitter = [\"best\", \"random\"]\n",
    "max_depth = [2, 3, 5, 10, 20]\n",
    "min_samples_leaf = [5, 10, 20, 50, 100]\n",
    "\n",
    "\n",
    "#Convert to dictionary\n",
    "hyperparameters = dict(criterion=criterion, splitter=splitter, max_depth=max_depth, min_samples_leaf=min_samples_leaf)\n",
    "\n",
    "print(hyperparameters)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 100 candidates, totalling 1000 fits\n"
     ]
    }
   ],
   "source": [
    "# estimator = ML model to be tested\n",
    "# param_grid = dictionary of hyperparameters to be tested\n",
    "# n_jobs = number of cores to be used in parallel, -1 means use all\n",
    "# cv = number of cross validation set elements\n",
    "# verbose = setting it to 2 will print updates to the terminal\n",
    "\n",
    "best_model = GridSearchCV(estimator=DecisionTreeClassifier(),\n",
    "                         param_grid=hyperparameters,\n",
    "                         n_jobs=-1,\n",
    "                         cv=10,\n",
    "                         verbose=2)\n",
    "\n",
    "# Fit the model\n",
    "best_model = best_model.fit(X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best criterion: entropy\n",
      "Best splitter: random\n",
      "Best max_depth: 20\n",
      "Best min_samples_leaf: 5\n"
     ]
    }
   ],
   "source": [
    "criterion = best_model.best_estimator_.get_params()['criterion']\n",
    "splitter = best_model.best_estimator_.get_params()['splitter']\n",
    "max_depth = best_model.best_estimator_.get_params()['max_depth']\n",
    "min_samples_leaf = best_model.best_estimator_.get_params()['min_samples_leaf']\n",
    "\n",
    "print(f'Best criterion: {criterion}')\n",
    "print(f'Best splitter: {splitter}')\n",
    "print(f'Best max_depth: {max_depth}')\n",
    "print(f'Best min_samples_leaf: {min_samples_leaf}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# model\n",
    "decision_tree = DecisionTreeClassifier(criterion=criterion,\n",
    "                                       splitter=splitter,\n",
    "                                       max_depth=max_depth,\n",
    "                                       min_samples_leaf=min_samples_leaf)\n",
    "\n",
    "# fit the model\n",
    "decision_tree.fit(X_train, y_train)\n",
    "\n",
    "# prediction\n",
    "predictions = decision_tree.predict(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Visualizing the Confusion Matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(y_test, predictions)\n",
    "plt.figure()\n",
    "sns.heatmap(cm, annot=True, fmt=\".3f\", linewidths=.5, square = True, cmap = 'Purples_r');\n",
    "plt.ylabel('Actual label');\n",
    "plt.xlabel('Predicted label');\n",
    "plt.title('Decision Tree Confusion Matrix', size = 15);\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision (macro average)--> 80.93%\n",
      "Precision (weighted average)--> 81.77%\n",
      "Recall (macro average)--> 81.16%\n",
      "Recall (weighted average)--> 80.4%\n",
      "F1 score (macro average)--> 80.39%\n",
      "F1 score (weighted average)--> 80.43%\n"
     ]
    }
   ],
   "source": [
    "\n",
    "precision_macro_average = precision_score(y_test, predictions, average='macro')\n",
    "precision_weighted_average = precision_score(y_test, predictions, average='weighted')\n",
    "\n",
    "recall_macro_average = recall_score(y_test, predictions, average='macro')\n",
    "recall_weighted_average = recall_score(y_test, predictions, average='weighted')\n",
    "\n",
    "f1_macro_average = f1_score(y_test, predictions, average='macro')\n",
    "f1_weighted_average = f1_score(y_test, predictions, average='weighted')\n",
    "\n",
    "\n",
    "print(f'Precision (macro average)--> {round(precision_macro_average*100, 2)}%')\n",
    "print(f'Precision (weighted average)--> {round(precision_weighted_average*100, 2)}%')\n",
    "\n",
    "print(f'Recall (macro average)--> {round(recall_macro_average*100, 2)}%')\n",
    "print(f'Recall (weighted average)--> {round(recall_weighted_average*100, 2)}%')\n",
    "\n",
    "print(f'F1 score (macro average)--> {round(f1_macro_average*100, 2)}%')\n",
    "print(f'F1 score (weighted average)--> {round(f1_weighted_average*100, 2)}%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.73      0.88      0.80       134\n",
      "           1       0.89      0.74      0.81       167\n",
      "\n",
      "    accuracy                           0.80       301\n",
      "   macro avg       0.81      0.81      0.80       301\n",
      "weighted avg       0.82      0.80      0.80       301\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### This model was Good"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Serialize f1 scores into a json file for later comparison between all models\n",
    "model_name = \"decision_tree\"\n",
    "entry = {\"f1_macro_average\":f1_macro_average, \"f1_weighted_average\":f1_weighted_average}\n",
    "\n",
    "\n",
    "with open(\"f1_scores.json\", \"r\") as file:\n",
    "    data = json.load(file)\n",
    "data[model_name] = entry\n",
    "with open(\"f1_scores.json\", \"w\") as file:\n",
    "    json.dump(data, file, indent=4)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "recsysspotify",
   "language": "python",
   "name": "recsysspotify"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}