Modelo2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "03793af6",
   "metadata": {},
   "source": [
    "# Multi-Label Classification\n",
    "### Random Forest Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "3931a250",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "import sklearn\n",
    "from sklearn import metrics, model_selection, tree\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.model_selection import RepeatedKFold\n",
    "from sklearn.model_selection import KFold\n",
    "from statistics import mean\n",
    "from sklearn.preprocessing import label_binarize"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8c81165b",
   "metadata": {},
   "source": [
    "#### Carga de data sets\n",
    "https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html\n",
    "https://www.analyticsvidhya.com/blog/2017/08/introduction-to-multi-label-classification/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "bfbae479",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Train data\n",
    "X_train = pd.read_csv('x_train.csv').to_numpy()\n",
    "y_train = pd.read_csv('y_train.csv')\n",
    "y_train = y_train['LABEL']\n",
    "\n",
    "# Test data\n",
    "X_test = pd.read_csv('x_test.csv').to_numpy()\n",
    "y_test = pd.read_csv('y_test.csv')\n",
    "y_test = y_test['LABEL']\n",
    "\n",
    "# Validation data\n",
    "X_validation = pd.read_csv('x_validation.csv').to_numpy()\n",
    "y_validation = pd.read_csv('y_validation.csv')\n",
    "y_validation = y_validation['LABEL']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "e6b096ae",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "f2dd48d4",
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = RandomForestClassifier(max_depth=3, random_state=6, n_jobs=10, n_estimators=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "55ae0514",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(max_depth=3, n_jobs=10, random_state=6)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "66d73045",
   "metadata": {},
   "source": [
    "#### Validation Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "af9aa0f0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9741294666661764"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.score(X_validation, y_validation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "15358cb9",
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_validation = clf.predict(X_validation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "12e071e2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9741294666661764\n",
      "Matriz de confusion Random Forest Classifier\n",
      " [[61498   997  4687  1283]\n",
      " [   67 64894     0     0]\n",
      " [    1     0 92343     0]\n",
      " [    0     0     0 46161]]\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      0.90      0.95     68465\n",
      "           1       0.98      1.00      0.99     64961\n",
      "           2       0.95      1.00      0.98     92344\n",
      "           3       0.97      1.00      0.99     46161\n",
      "\n",
      "    accuracy                           0.97    271931\n",
      "   macro avg       0.98      0.97      0.97    271931\n",
      "weighted avg       0.98      0.97      0.97    271931\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(metrics.accuracy_score(y_validation, pred_validation))\n",
    "print('Matriz de confusion Random Forest Classifier\\n', metrics.confusion_matrix(y_validation, pred_validation))\n",
    "print(metrics.classification_report(y_validation, pred_validation, target_names=['0', '1', '2', '3']))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "45c96647",
   "metadata": {},
   "source": [
    "#### Test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "0ae5272e",
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "d3ab2d0f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9745988045643905"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "254c8ef2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9745988045643905\n",
      "Matriz de confusion Random Forest Classifier\n",
      " [[124669   1941   9278   2661]\n",
      " [   139 133153      1      0]\n",
      " [     4      0 187443      0]\n",
      " [     0      0      0  92811]]\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      0.90      0.95    138549\n",
      "           1       0.99      1.00      0.99    133293\n",
      "           2       0.95      1.00      0.98    187447\n",
      "           3       0.97      1.00      0.99     92811\n",
      "\n",
      "    accuracy                           0.97    552100\n",
      "   macro avg       0.98      0.97      0.98    552100\n",
      "weighted avg       0.98      0.97      0.97    552100\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(metrics.accuracy_score(y_test, predictions))\n",
    "print('Matriz de confusion Random Forest Classifier\\n', metrics.confusion_matrix(y_test, predictions))\n",
    "print(metrics.classification_report(y_test, predictions, target_names=['0', '1', '2', '3']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "d1c1bace",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "e36753db",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['modelo2.h5']"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(clf, 'modelo2.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "84e92d78",
   "metadata": {},
   "source": [
    "## Grafica ROC\n",
    "Extraído de: https://laurenliz22.github.io/roc_curve_multiclass_predictions_random_forest_classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "3670d156",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Calculate the y_score\n",
    "y_score = clf.predict_proba(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "34516c3b",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Binarize the output\n",
    "y_test_bin = label_binarize(y_test, classes=[0, 1, 2, 3])\n",
    "n_classes = y_test_bin.shape[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "10975e70",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.rcParams['figure.figsize'] = (16, 10)\n",
    "fpr = dict()\n",
    "tpr = dict()\n",
    "roc_auc = dict()\n",
    "\n",
    "for i in range(n_classes):\n",
    "  fpr[i], tpr[i], _ = metrics.roc_curve(y_test_bin[:, i], y_score[:, i])\n",
    "  plt.plot(fpr[i], tpr[i], lw=2, label='ROC curve (area = %0.2f) for label %i' % (metrics.auc(fpr[i], tpr[i]), i))\n",
    "\n",
    "plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.legend(loc=\"best\")\n",
    "plt.grid(alpha=.4)\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title('Receiver Operating Characteristic Curves')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "665bf727",
   "metadata": {},
   "source": [
    "## Cross Validation\n",
    "Fuente de consulta: https://machinelearningmastery.com/repeated-k-fold-cross-validation-with-python/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "262a46e6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.97336047 0.97434344 0.97274487 0.97404557 0.97549521 0.97372785\n",
      " 0.97421437 0.97182148 0.97431366 0.97426376]\n"
     ]
    }
   ],
   "source": [
    "rf_cv_score=cross_val_score(estimator=clf, X=X_train, y=y_train, cv=10, n_jobs=-1)\n",
    "print(rf_cv_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "754bffdd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# prepare the cross-validation procedure\n",
    "cv = KFold(n_splits=10, random_state=1, shuffle=True)\n",
    "# evaluate model\n",
    "scores = cross_val_score(clf, X_train, y_train, scoring='accuracy', cv=cv, n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5b917df4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.975 (0.001)\n"
     ]
    }
   ],
   "source": [
    "# report performance\n",
    "print('Accuracy: %.3f (%.3f)' % (np.mean(scores), np.std(scores)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "fce6bc46",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train: [      0       1       3 ... 1007146 1007147 1007148] Validation: [      2       6      17 ... 1007091 1007125 1007130]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     13      23      32 ... 1007108 1007112 1007119]\n",
      "Train: [      0       1       2 ... 1007145 1007146 1007148] Validation: [     14      19      24 ... 1007128 1007137 1007147]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      7      12      20 ... 1007134 1007139 1007144]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     31      66      73 ... 1007126 1007142 1007143]\n",
      "Train: [      1       2       3 ... 1007146 1007147 1007148] Validation: [      0      16      25 ... 1007092 1007096 1007131]\n",
      "Train: [      0       1       2 ... 1007144 1007145 1007147] Validation: [      4       8       9 ... 1007141 1007146 1007148]\n",
      "Train: [      0       2       4 ... 1007146 1007147 1007148] Validation: [      1       3      10 ... 1007132 1007136 1007138]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     27      49      52 ... 1007122 1007127 1007145]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      5      39      58 ... 1007081 1007099 1007103]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      3      32      41 ... 1007123 1007141 1007143]\n",
      "Train: [      1       2       3 ... 1007146 1007147 1007148] Validation: [      0       7      20 ... 1007128 1007130 1007137]\n",
      "Train: [      0       1       3 ... 1007146 1007147 1007148] Validation: [      2       6       9 ... 1007110 1007116 1007127]\n",
      "Train: [      0       1       2 ... 1007145 1007147 1007148] Validation: [     17      21      22 ... 1007140 1007144 1007146]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      4       8      12 ... 1007135 1007136 1007139]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     11      34      40 ... 1007124 1007142 1007145]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     15      19      28 ... 1007098 1007105 1007125]\n",
      "Train: [      0       1       2 ... 1007145 1007146 1007148] Validation: [      5      10      16 ... 1007117 1007118 1007147]\n",
      "Train: [      0       1       2 ... 1007145 1007146 1007147] Validation: [     13      23      27 ... 1007062 1007133 1007148]\n",
      "Train: [      0       2       3 ... 1007146 1007147 1007148] Validation: [      1      14      37 ... 1007115 1007119 1007138]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      4      30      40 ... 1007131 1007140 1007144]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      9      13      15 ... 1007091 1007105 1007111]\n",
      "Train: [      2       3       4 ... 1007145 1007146 1007147] Validation: [      0       1       7 ... 1007138 1007143 1007148]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     34      46      60 ... 1007137 1007141 1007142]\n",
      "Train: [      0       1       2 ... 1007145 1007147 1007148] Validation: [     19      27      45 ... 1007119 1007133 1007146]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      5      10      14 ... 1007125 1007130 1007136]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     11      51      53 ... 1007068 1007086 1007123]\n",
      "Train: [      0       1       2 ... 1007144 1007146 1007148] Validation: [      6      18      23 ... 1007139 1007145 1007147]\n",
      "Train: [      0       1       3 ... 1007146 1007147 1007148] Validation: [      2      20      29 ... 1007108 1007110 1007126]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      3      12      21 ... 1007095 1007128 1007134]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      7      18      69 ... 1007093 1007119 1007138]\n",
      "Train: [      0       1       2 ... 1007142 1007143 1007146] Validation: [     11      25      67 ... 1007145 1007147 1007148]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      4       9      19 ... 1007072 1007128 1007131]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     17      20      33 ... 1007132 1007135 1007137]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     23      24      40 ... 1007096 1007124 1007140]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     10      12      14 ... 1007126 1007130 1007133]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      3       6      26 ... 1007129 1007134 1007136]\n",
      "Train: [      0       2       3 ... 1007145 1007147 1007148] Validation: [      1      15      34 ... 1007141 1007143 1007146]\n",
      "Train: [      1       2       3 ... 1007146 1007147 1007148] Validation: [      0       5      13 ... 1007113 1007118 1007142]\n",
      "Train: [      0       1       3 ... 1007146 1007147 1007148] Validation: [      2       8      27 ... 1007112 1007114 1007122]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      9      14      20 ... 1007104 1007105 1007135]\n",
      "Train: [      0       1       2 ... 1007145 1007146 1007148] Validation: [     16      19      25 ... 1007142 1007143 1007147]\n",
      "Train: [      1       2       3 ... 1007144 1007145 1007147] Validation: [      0       8      10 ... 1007132 1007146 1007148]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     55      57      58 ... 1007138 1007139 1007140]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     11      12      24 ... 1007121 1007123 1007144]\n",
      "Train: [      0       1       3 ... 1007146 1007147 1007148] Validation: [      2      18      28 ... 1007094 1007095 1007128]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      3       6      13 ... 1007133 1007134 1007136]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [     22      23      38 ... 1007069 1007112 1007145]\n",
      "Train: [      0       1       2 ... 1007146 1007147 1007148] Validation: [      4       5       7 ... 1007111 1007120 1007125]\n",
      "Train: [      0       2       3 ... 1007146 1007147 1007148] Validation: [      1      21      34 ... 1007103 1007127 1007141]\n"
     ]
    }
   ],
   "source": [
    "kf = RepeatedKFold(n_splits=10, n_repeats=5, random_state=None) \n",
    "\n",
    "for train_index, test_index in kf.split(X_train):\n",
    "      print(\"Train:\", train_index, \"Validation:\",test_index)\n",
    "      x_train, x_test = X_train[train_index], X_train[test_index] \n",
    "      Y_train, Y_test = y_train[train_index], y_train[test_index]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "a9958826",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 100715  100716  100717 ... 1007146 1007147 1007148] [     0      1      2 ... 100712 100713 100714]\n",
      "[      0       1       2 ... 1007146 1007147 1007148] [100715 100716 100717 ... 201427 201428 201429]\n",
      "[      0       1       2 ... 1007146 1007147 1007148] [201430 201431 201432 ... 302142 302143 302144]\n",
      "[      0       1       2 ... 1007146 1007147 1007148] [302145 302146 302147 ... 402857 402858 402859]\n",
      "[      0       1       2 ... 1007146 1007147 1007148] [402860 402861 402862 ... 503572 503573 503574]\n",
      "[      0       1       2 ... 1007146 1007147 1007148] [503575 503576 503577 ... 604287 604288 604289]\n",
      "[      0       1       2 ... 1007146 1007147 1007148] [604290 604291 604292 ... 705002 705003 705004]\n",
      "[      0       1       2 ... 1007146 1007147 1007148] [705005 705006 705007 ... 805717 805718 805719]\n",
      "[      0       1       2 ... 1007146 1007147 1007148] [805720 805721 805722 ... 906432 906433 906434]\n",
      "[     0      1      2 ... 906432 906433 906434] [ 906435  906436  906437 ... 1007146 1007147 1007148]\n"
     ]
    }
   ],
   "source": [
    "kf = KFold(n_splits=10)\n",
    "for train, test in kf.split(X_train):\n",
    "    print(\"%s %s\" % (train, test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e2bdc4e7",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}