20-[TALLER]_Pipeline_Solucion_Problema_de_ML.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<a href=\"https://colab.research.google.com/github/UDEA-Esp-Analitica-y-Ciencia-de-Datos/EACD-04-MACHINE-LEARNING-1/blob/master/20-[TALLER]_Pipeline_Solucion_Problema_de_ML.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\n",
    "\n",
    "\n",
    "**Recuerda que una vez abierto, Da clic en \"Copiar en Drive\", de lo contrario no podras alamancenar tu progreso**\n",
    "\n",
    "Nota: no olvide ir ejecutando las celdas de código de arriba hacia abajo para que no tenga errores de importación de librerías o por falta de definición de variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "!wget --no-cache -O init.py -q https://raw.githubusercontent.com/UDEA-Esp-Analitica-y-Ciencia-de-Datos/EACD-04-MACHINE-LEARNING-1/master/init.py\n",
    "import init; init.init(force_download=False); "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Taller\n",
    "\n",
    "El objetivo este taller es recoger lo visto en el módulo y aplicarlo a la solución de un problema real. El problema que vamos a resolver es la identificación y clasificación de comentarios tóxicos en línea. Los detalles del conjunto de datos pueden ser consulatos [aquí](https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge/overview)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install scikit-learn==0.24"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.model_selection import KFold\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.multiclass import OneVsRestClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.metrics import accuracy_score, make_scorer\n",
    "from sklearn.metrics import multilabel_confusion_matrix\n",
    "from sklearn.metrics._plot.confusion_matrix import ConfusionMatrixDisplay\n",
    "from sklearn.metrics import hamming_loss\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "from sklearn.utils.class_weight import compute_class_weight"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('local/data/train.csv.zip', compression='zip', header=0, sep=',', quotechar='\"')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Veamos algunas muestras de la base datos:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0000997932d777bf</td>\n",
       "      <td>explanation\\nwhy the edits made under my usern...</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>1</th>\n",
       "      <td>000103f0d9cfb60f</td>\n",
       "      <td>d'aww! he matches this background colour i'm s...</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>2</th>\n",
       "      <td>000113f07ec002fd</td>\n",
       "      <td>hey man, i'm really not trying to edit war. it...</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>3</th>\n",
       "      <td>0001b41b1c6bb37e</td>\n",
       "      <td>\"\\nmore\\ni can't make any real suggestions on ...</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>4</th>\n",
       "      <td>0001d958c54c6e35</td>\n",
       "      <td>you, sir, are my hero. any chance you remember...</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>5</th>\n",
       "      <td>00025465d4725e87</td>\n",
       "      <td>\"\\n\\ncongratulations from me as well, use the ...</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>6</th>\n",
       "      <td>0002bcb3da6cb337</td>\n",
       "      <td>cocksucker before you piss around on my work</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>00031b1e95af7921</td>\n",
       "      <td>your vandalism to the matt shirvington article...</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>8</th>\n",
       "      <td>00037261f536c51d</td>\n",
       "      <td>sorry if the word 'nonsense' was offensive to ...</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>9</th>\n",
       "      <td>00040093b2687caa</td>\n",
       "      <td>alignment on this subject and which are contra...</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",
       "</div>"
      ],
      "text/plain": [
       "                 id                                       comment_text  toxic  \\\n",
       "0  0000997932d777bf  explanation\\nwhy the edits made under my usern...      0   \n",
       "1  000103f0d9cfb60f  d'aww! he matches this background colour i'm s...      0   \n",
       "2  000113f07ec002fd  hey man, i'm really not trying to edit war. it...      0   \n",
       "3  0001b41b1c6bb37e  \"\\nmore\\ni can't make any real suggestions on ...      0   \n",
       "4  0001d958c54c6e35  you, sir, are my hero. any chance you remember...      0   \n",
       "5  00025465d4725e87  \"\\n\\ncongratulations from me as well, use the ...      0   \n",
       "6  0002bcb3da6cb337       cocksucker before you piss around on my work      1   \n",
       "7  00031b1e95af7921  your vandalism to the matt shirvington article...      0   \n",
       "8  00037261f536c51d  sorry if the word 'nonsense' was offensive to ...      0   \n",
       "9  00040093b2687caa  alignment on this subject and which are contra...      0   \n",
       "\n",
       "   severe_toxic  obscene  threat  insult  identity_hate  \n",
       "0             0        0       0       0              0  \n",
       "1             0        0       0       0              0  \n",
       "2             0        0       0       0              0  \n",
       "3             0        0       0       0              0  \n",
       "4             0        0       0       0              0  \n",
       "5             0        0       0       0              0  \n",
       "6             1        1       0       1              0  \n",
       "7             0        0       0       0              0  \n",
       "8             0        0       0       0              0  \n",
       "9             0        0       0       0              0  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"comment_text\"] = df[\"comment_text\"].str.lower()\n",
    "df[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(159571, 8)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Desde el punto de vista de ML, el problema predictivo corresponde a un problema de múltiples etiquetas:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax= df.hist(column=['toxic','severe_toxic','obscene','threat','insult','identity_hate'],sharey=True,figsize=(10, 10))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "La base de datos completa es muy grande, por lo tanto tomaremos sólo una parte para este ejercicio. Abordaremos el problema de predecir tres clases de contenido: 'toxic', 'obscene','insult'. Adicionalmente, teniendo en cuenta que muchas muestras no pertenecen a ninguna de las clases, eliminaremos principalmente muestras de ese conjunto."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(seed=0)\n",
    "df2 = df[(df['toxic']==0) & (df['severe_toxic']==0) & (df['obscene']==0) & (df['threat']==0) & (df['insult']==0) & (df['identity_hate']==0)]\n",
    "df = df.drop(['severe_toxic', 'threat', 'identity_hate'], axis=1)\n",
    "df2 = df2.drop(['severe_toxic', 'threat', 'identity_hate'], axis=1)\n",
    "df1 = df[(df['toxic']!=0) | (df['obscene']!=0) | (df['insult']!=0)]\n",
    "remove_n = 130000\n",
    "drop_indices = np.random.choice(df2.index, remove_n, replace=False)\n",
    "df2 = df2.drop(drop_indices)\n",
    "remove_n = 5000\n",
    "drop_indices = np.random.choice(df1.index, remove_n, replace=False)\n",
    "df1 = df1.drop(drop_indices)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_final = pd.concat([df1, df2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(24495, 5)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_final.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Algunas muestras pueden no tener texto asociado, así que las imputaremos con la palabra: unknown "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "COMMENT = 'comment_text'\n",
    "df_final[COMMENT].fillna(\"unknown\", inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'=tony sidaway is obviously a fistfuckee. he loves an arm up his ass.'"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_final['comment_text'][df_final.index[6]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(358.48169014084505, 605.5713565714727, 5000)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lens = df_final.comment_text.str.len()\n",
    "lens.mean(), lens.std(), lens.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lens.hist();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Nota**: De las anteriores gráficas se puede observar que todas las clases a predecir están desbalanceadas, pero la estrategia de validación estratificada estándar sólo puede fijarse en una. Un método al ternativo, llamado estratificación iterativa puede consultarse en el siguiente [artículo](https://link.springer.com/content/pdf/10.1007%2F978-3-642-23808-6_10.pdf) Su implementación está disponible en la librería [scikit-multilearn](http://scikit.ml/stratification.html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "import re, string\n",
    "re_tok = re.compile(f'([{string.punctuation}“”¨«»®´·º½¾¿¡§£₤‘’])')\n",
    "def tokenize(s): return re_tok.sub(r' \\1 ', s).split()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(df_final['comment_text'],df_final[['toxic','obscene','insult']], test_size=0.2, random_state=42, stratify=df_final['insult'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Ejemplo: Random Forest en esquema One vs Rest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_unique = ['toxic','obscene','insult']\n",
    "classifier = Pipeline([\n",
    "    ('tfidf', TfidfVectorizer(ngram_range=(1,2), tokenizer=tokenize,\n",
    "               min_df=3, max_df=0.9, strip_accents='unicode', max_features = 2000, use_idf=1,\n",
    "               smooth_idf=1, sublinear_tf=1, stop_words='english')),\n",
    "    ('scaler', StandardScaler(with_mean=False)),\n",
    "    ('clf', OneVsRestClassifier(RandomForestClassifier(class_weight='balanced_subsample')))])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Parameters of pipelines can be set using ‘__’ separated parameter names:\n",
    "tuned_parameters = [{\"tfidf__max_features\":[1000,2000],\n",
    "                     \"clf__estimator__n_estimators\": [100,150,200,250,300,350,400],\n",
    "                     \"clf__estimator__max_depth\": [10,12,14,16],\n",
    "                     \"clf__estimator__max_features\":['auto','sqrt','log2']},\n",
    "                   {\"clf__estimator__n_estimators\": [100,150,200,250,300,350,400],\n",
    "                     \"clf__estimator__max_depth\": [None],\n",
    "                     \"clf__estimator__max_features\":['auto','sqrt','log2'],\n",
    "                     \"clf__estimator__min_samples_split\":[10,20,30,40]}]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 168 candidates, totalling 504 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 16 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   9 tasks      | elapsed:   13.9s\n",
      "[Parallel(n_jobs=-1)]: Done 130 tasks      | elapsed:  2.0min\n",
      "/home/julian/.local/lib/python3.7/site-packages/joblib/externals/loky/process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "[Parallel(n_jobs=-1)]: Done 333 tasks      | elapsed:  9.4min\n",
      "[Parallel(n_jobs=-1)]: Done 504 out of 504 | elapsed: 18.2min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=KFold(n_splits=3, random_state=None, shuffle=False),\n",
       "             error_score=nan,\n",
       "             estimator=Pipeline(memory=None,\n",
       "                                steps=[('tfidf',\n",
       "                                        TfidfVectorizer(analyzer='word',\n",
       "                                                        binary=False,\n",
       "                                                        decode_error='strict',\n",
       "                                                        dtype=<class 'numpy.float64'>,\n",
       "                                                        encoding='utf-8',\n",
       "                                                        input='content',\n",
       "                                                        lowercase=True,\n",
       "                                                        max_df=0.9,\n",
       "                                                        max_features=2000,\n",
       "                                                        min_df=3,\n",
       "                                                        ngram_range=(1, 2),\n",
       "                                                        norm='l2',\n",
       "                                                        preprocesso...\n",
       "                          'clf__estimator__n_estimators': [100, 150, 200, 250,\n",
       "                                                           300, 350, 400]},\n",
       "                         {'clf__estimator__max_depth': [None],\n",
       "                          'clf__estimator__max_features': ['auto', 'sqrt',\n",
       "                                                           'log2'],\n",
       "                          'clf__estimator__min_samples_split': [10, 20, 30, 40],\n",
       "                          'clf__estimator__n_estimators': [100, 150, 200, 250,\n",
       "                                                           300, 350, 400]}],\n",
       "             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,\n",
       "             scoring=make_scorer(accuracy_score), verbose=2)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "kf = KFold(n_splits=3)\n",
    "model_tunning = GridSearchCV(classifier, param_grid=tuned_parameters,\n",
    "                             scoring=make_scorer(accuracy_score), n_jobs=-1, cv=kf, verbose=2)\n",
    "\n",
    "model_tunning.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('clf', OneVsRestClassifier(estimator=RandomForestClassifier(bootstrap=True,\n",
       "                                                      ccp_alpha=0.0,\n",
       "                                                      class_weight='balanced_subsample',\n",
       "                                                      criterion='gini',\n",
       "                                                      max_depth=None,\n",
       "                                                      max_features='auto',\n",
       "                                                      max_leaf_nodes=None,\n",
       "                                                      max_samples=None,\n",
       "                                                      min_impurity_decrease=0.0,\n",
       "                                                      min_impurity_split=None,\n",
       "                                                      min_samples_leaf=1,\n",
       "                                                      min_samples_split=20,\n",
       "                                                      min_weight_fraction_leaf=0.0,\n",
       "                                                      n_estimators=250,\n",
       "                                                      n_jobs=None,\n",
       "                                                      oob_score=False,\n",
       "                                                      random_state=None,\n",
       "                                                      verbose=0,\n",
       "                                                      warm_start=False),\n",
       "                     n_jobs=None))"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_tunning.best_estimator_.steps[2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6964686670749133"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_tunning.best_score_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "87939     \"for instance, read what the dynasts were sayi...\n",
       "45721     \"\\n\\nthe publication \"\"the chamcha age\"\" - an ...\n",
       "42610     don't be stupid. you know, i'm twice as irrita...\n",
       "127722    whoever created this page is either very jelou...\n",
       "135802    that's libel right there. murder is defined as...\n",
       "                                ...                        \n",
       "58193     \"\\nthank you for applying to access the accoun...\n",
       "159460    , 24 january 2010 (utc)\\nwikipedia is hosted i...\n",
       "4251      that's a good idea (assuming someone from the ...\n",
       "16255     \"\\nactually, we've been removing many \"\"terror...\n",
       "50872     \"\\n\\n please do not vandalize pages, as you di...\n",
       "Name: comment_text, Length: 4899, dtype: object"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = model_tunning.predict(X_test)\n",
    "cm = multilabel_confusion_matrix(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axes = plt.subplots(1, 3, figsize=(15,10))\n",
    "\n",
    "for i, (k, ax) in enumerate(zip(y_unique, axes.flatten())):\n",
    "    label = []\n",
    "    label.append(k)\n",
    "    label.append('no '+ k) \n",
    "    disp = ConfusionMatrixDisplay(confusion_matrix=(cm[i].T/np.sum(cm[i],axis=1)).T, display_labels=label)\n",
    "    disp.plot(include_values=True,cmap=plt.cm.Blues, ax=ax, xticks_rotation='horizontal',\n",
    "          values_format=None)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Global train error = 0.013676260461318637\n",
      "Global test error = 0.13213581002925767\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = model_tunning.predict(X_train)\n",
    "print(f\"Global train error = {hamming_loss(y_train, y_pred_train)}\")\n",
    "print(f\"Global test error = {hamming_loss(y_test, y_pred)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Podemos averiguar cuántas palabras fueron incluidas en el diccionario:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2000"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_tunning.best_estimator_.steps[2][1].estimators_[0].n_features_"
   ]
  },
  {
   "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": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "p_pred = model_tunning.predict_proba(X_test)\n",
    "for i,j in enumerate(y_unique):\n",
    "    score_pred = p_pred[:,i].copy()\n",
    "    fpr, tpr, _ = roc_curve(y_test.values[:,i], score_pred)\n",
    "    plt.plot(fpr, tpr, label=j+' '+ '(AUC=' + str(round(auc(fpr, tpr),2)) +')')\n",
    "plt.legend()\n",
    "plt.title('Curvas ROC para cada etiqueta')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Variante: Random Forest Multi-task"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "class_weight_list = []\n",
    "for i in range(len(y_unique)):\n",
    "    labels = np.unique(y_train.values[:,i])\n",
    "    temp = compute_class_weight('balanced',labels,y_train.values[:,i])\n",
    "    class_weight_list.append({labels[0]:temp[0], labels[1]:temp[1]})\n",
    "\n",
    "classifier = Pipeline([\n",
    "    ('tfidf', TfidfVectorizer(ngram_range=(1,2), tokenizer=tokenize,\n",
    "               min_df=3, max_df=0.9, strip_accents='unicode', max_features = 2000, use_idf=1,\n",
    "               smooth_idf=1, sublinear_tf=1, stop_words='english')),\n",
    "    ('scaler', StandardScaler(with_mean=False)),\n",
    "    ('clf', RandomForestClassifier(class_weight=class_weight_list))])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Parameters of pipelines can be set using ‘__’ separated parameter names:\n",
    "tuned_parameters = [{\"clf__n_estimators\": [100,150,200,250,300,350,400],\n",
    "                     \"clf__max_depth\": [10,12,14,16],\n",
    "                     \"clf__max_features\":['auto','sqrt','log2']},\n",
    "                   {\"clf__n_estimators\": [100,150,200,250,300,350,400],\n",
    "                     \"clf__max_depth\": [None],\n",
    "                     \"clf__max_features\":['auto','sqrt','log2'],\n",
    "                     \"clf__min_samples_split\":[10,20,30,40]}]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 168 candidates, totalling 504 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 16 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   9 tasks      | elapsed:    7.9s\n",
      "[Parallel(n_jobs=-1)]: Done 130 tasks      | elapsed:  1.1min\n",
      "/home/julian/.local/lib/python3.7/site-packages/joblib/externals/loky/process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "[Parallel(n_jobs=-1)]: Done 333 tasks      | elapsed:  4.9min\n",
      "[Parallel(n_jobs=-1)]: Done 504 out of 504 | elapsed:  9.0min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=KFold(n_splits=3, random_state=None, shuffle=False),\n",
       "             error_score=nan,\n",
       "             estimator=Pipeline(memory=None,\n",
       "                                steps=[('tfidf',\n",
       "                                        TfidfVectorizer(analyzer='word',\n",
       "                                                        binary=False,\n",
       "                                                        decode_error='strict',\n",
       "                                                        dtype=<class 'numpy.float64'>,\n",
       "                                                        encoding='utf-8',\n",
       "                                                        input='content',\n",
       "                                                        lowercase=True,\n",
       "                                                        max_df=0.9,\n",
       "                                                        max_features=2000,\n",
       "                                                        min_df=3,\n",
       "                                                        ngram_range=(1, 2),\n",
       "                                                        norm='l2',\n",
       "                                                        preprocesso...\n",
       "                          'clf__max_features': ['auto', 'sqrt', 'log2'],\n",
       "                          'clf__n_estimators': [100, 150, 200, 250, 300, 350,\n",
       "                                                400]},\n",
       "                         {'clf__max_depth': [None],\n",
       "                          'clf__max_features': ['auto', 'sqrt', 'log2'],\n",
       "                          'clf__min_samples_split': [10, 20, 30, 40],\n",
       "                          'clf__n_estimators': [100, 150, 200, 250, 300, 350,\n",
       "                                                400]}],\n",
       "             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,\n",
       "             scoring=make_scorer(accuracy_score), verbose=2)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "kf = KFold(n_splits=3)\n",
    "model_tunning = GridSearchCV(classifier, param_grid=tuned_parameters,\n",
    "                             scoring=make_scorer(accuracy_score), n_jobs=-1, cv=kf, verbose=2)\n",
    "\n",
    "model_tunning.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, ccp_alpha=0.0,\n",
       "                       class_weight=[{0: 0.8792175161521896,\n",
       "                                      1: 1.1592522479886418},\n",
       "                                     {0: 0.6563504823151125,\n",
       "                                      1: 2.0989717223650386},\n",
       "                                     {0: 0.6429977687360546,\n",
       "                                      1: 2.2482790270766406}],\n",
       "                       criterion='gini', max_depth=None, max_features='sqrt',\n",
       "                       max_leaf_nodes=None, max_samples=None,\n",
       "                       min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                       min_samples_leaf=1, min_samples_split=10,\n",
       "                       min_weight_fraction_leaf=0.0, n_estimators=300,\n",
       "                       n_jobs=None, oob_score=False, random_state=None,\n",
       "                       verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_tunning.best_estimator_.steps[2][1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6730965503163912"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_tunning.best_score_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = model_tunning.predict(X_test)\n",
    "cm = multilabel_confusion_matrix(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axes = plt.subplots(1, 3, figsize=(15,10))\n",
    "\n",
    "for i, (k, ax) in enumerate(zip(y_unique, axes.flatten())):\n",
    "    label = []\n",
    "    label.append(k)\n",
    "    label.append('no '+ k) \n",
    "    disp = ConfusionMatrixDisplay(confusion_matrix=(cm[i].T/np.sum(cm[i],axis=1)).T, display_labels=label)\n",
    "    disp.plot(include_values=True,cmap=plt.cm.Blues, ax=ax, xticks_rotation='horizontal',\n",
    "          values_format=None)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Global train error = 0.013438116622439954\n",
      "Global test error = 0.1493502075253453\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = model_tunning.predict(X_train)\n",
    "print(f\"Global train error = {hamming_loss(y_train, y_pred_train)}\")\n",
    "print(f\"Global test error = {hamming_loss(y_test, y_pred)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Ejercicio 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Ejercicio de código\n",
    "p_pred = model_tunning.predict_proba(X_test)\n",
    "for i,j in enumerate(y_unique):\n",
    "    #OJO: Son tres salidas...\n",
    "    score_pred = ...\n",
    "    fpr, tpr, _ = roc_curve(y_test.values[:,i], score_pred)\n",
    "    plt.plot(fpr, tpr, label=j+' '+ '(AUC=' + str(round(auc(fpr, tpr),2)) +')')\n",
    "plt.legend()\n",
    "plt.title('Curvas ROC para cada etiqueta')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Ejericicio 2.\n",
    "\n",
    "Realizar un proceso similar al anterior pero para una Red Neuronal Aritificial. En este caso evalúe máximo 2 capas ocultas con las siguientes configuraciones: (64,32) (128,64). Incluya parámetro de regularización dentro de la búsqueda de hiperparámetros y modifique también el número máximo de época de acuerdo con el siguiente conjunto: [30, 60, 100]. También puede variar el tamaño del batch: batch_size y la política de la tasa de aprendizaje: learning_rate.\n",
    "\n",
    "Recuerde que la RNA soporta la estrategia de early_stopping, para la que debe definir además el parámetro n_iter_no_change y validation_fraction."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Ejercicio de código\n",
    "\n",
    "\n",
    "..."
   ]
  }
 ],
 "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": "3.7.6"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}