Clasificacion.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Clasificación y métricas de evaluación\n",
    "**Autor:** José A. Troyano, Beatriz Pontes &nbsp;&nbsp;&nbsp; **Última modificación:** 16/01/2019"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Contenido\n",
    "\n",
    "1. <a href=\"#clasificador\">Entrenamiento de un clasificador</a><br>\n",
    "    1.1 <a href=\"#logistica\">Regresión logística</a><br>\n",
    "    1.2 <a href=\"#estimadores\">Estimadores de sklearn: _fit_, _predict_ y _predict_\\__proba_</a>\n",
    "2. <a href=\"#evaluacion_tt\">Evaluación train/test</a><br>\n",
    "    2.1 <a href=\"#separacion\">Separación entrenamiento y validación</a> <br>\n",
    "    2.2 <a href=\"#accuracy\">Cálculo del accuracy</a><br>\n",
    "    2.3 <a href=\"#pertenencia\">Probabilidades de pertenencia</a><br>\n",
    "3. <a href=\"#evaluacion_cruzada\">Evaluación mediante validación cruzada</a> <br>\n",
    "4. <a href=\"#matriz_cf\">Matriz de confusión y otras métricas</a><br>\n",
    "    4.1 <a href=\"#aciertos_fallos\">Número de aciertos y fallos: TP, FP, TN, FN</a><br>\n",
    "    4.2 <a href=\"#medidas\">accuracy, precission, recall y f1</a> <br>\n",
    "    4.3 <a href=\"#rates\">_True Positive Rate_ y _False Positive Rate_</a> <br>\n",
    "5. <a href=\"#curvas\">Curva ROC y curva Precision-Recall</a> <br>\n",
    "6. <a href=\"#otros_clasif\">Otros clasificadores</a> <br>\n",
    "    6.1 <a href=\"#arboles\">Árboles de Decisión</a> <br>\n",
    "    6.2 <a href=\"#randomF\">Random Forest</a> <br>\n",
    "------------------------------------------------------"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "En este notebook usaremos un dataset para entrenar un clasificador de Sklearn y veremos distintas formas de evaluar la calidad del clasificador obtenido. \n",
    "\n",
    "Empezaremos por importar todos los elementos que usaremos a lo largo del notebook:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting scikit-plot\n",
      "  Downloading https://files.pythonhosted.org/packages/7c/47/32520e259340c140a4ad27c1b97050dd3254fdc517b1d59974d47037510e/scikit_plot-0.3.7-py3-none-any.whl\n",
      "Requirement already satisfied: graphviz in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (0.8.4)\n",
      "Collecting pydotplus\n",
      "  Downloading https://files.pythonhosted.org/packages/60/bf/62567830b700d9f6930e9ab6831d6ba256f7b0b730acb37278b0ccdffacf/pydotplus-2.0.2.tar.gz (278kB)\n",
      "Requirement already satisfied: scipy>=0.9 in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from scikit-plot) (1.1.0)\n",
      "Requirement already satisfied: scikit-learn>=0.18 in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from scikit-plot) (0.20.0)\n",
      "Collecting joblib>=0.10 (from scikit-plot)\n",
      "  Downloading https://files.pythonhosted.org/packages/49/d9/4ea194a4c1d0148f9446054b9135f47218c23ccc6f649aeb09fab4c0925c/joblib-0.13.1-py2.py3-none-any.whl (278kB)\n",
      "Requirement already satisfied: matplotlib>=1.4.0 in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from scikit-plot) (3.0.1)\n",
      "Requirement already satisfied: pyparsing>=2.0.1 in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from pydotplus) (2.3.0)\n",
      "Requirement already satisfied: numpy>=1.8.2 in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from scikit-learn>=0.18->scikit-plot) (1.15.4)\n",
      "Requirement already satisfied: cycler>=0.10 in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from matplotlib>=1.4.0->scikit-plot) (0.10.0)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from matplotlib>=1.4.0->scikit-plot) (1.0.1)\n",
      "Requirement already satisfied: python-dateutil>=2.1 in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from matplotlib>=1.4.0->scikit-plot) (2.7.5)\n",
      "Requirement already satisfied: six in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from cycler>=0.10->matplotlib>=1.4.0->scikit-plot) (1.11.0)\n",
      "Requirement already satisfied: setuptools in c:\\users\\belladona\\anaconda3\\envs\\ztdl\\lib\\site-packages (from kiwisolver>=1.0.1->matplotlib>=1.4.0->scikit-plot) (39.1.0)\n",
      "Building wheels for collected packages: pydotplus\n",
      "  Running setup.py bdist_wheel for pydotplus: started\n",
      "  Running setup.py bdist_wheel for pydotplus: finished with status 'done'\n",
      "  Stored in directory: C:\\Users\\Belladona\\AppData\\Local\\pip\\Cache\\wheels\\35\\7b\\ab\\66fb7b2ac1f6df87475b09dc48e707b6e0de80a6d8444e3628\n",
      "Successfully built pydotplus\n",
      "Installing collected packages: joblib, scikit-plot, pydotplus\n",
      "Successfully installed joblib-0.13.1 pydotplus-2.0.2 scikit-plot-0.3.7\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "You are using pip version 18.1, however version 19.0.1 is available.\n",
      "You should consider upgrading via the 'python -m pip install --upgrade pip' command.\n"
     ]
    }
   ],
   "source": [
    "!pip install scikit-plot graphviz pydotplus"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Lectura de datos\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from sklearn.datasets import load_breast_cancer\n",
    "\n",
    "# Clasificación\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn import tree\n",
    "\n",
    "# Evaluación\n",
    "from sklearn import metrics\n",
    "from sklearn.model_selection import train_test_split, cross_val_predict, cross_val_score\n",
    "\n",
    "# Visualización\n",
    "import scikitplot as skplt\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from graphviz import Source\n",
    "import pydotplus\n",
    "\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "import scipy.stats"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Usaremos el dataset _breast cancer_, disponible en el repositorio UCI y también incluido en el conjunto de datasets de prueba de Sklearn. \n",
    "\n",
    "El dataset contiene 569 registros correspondientes a pacientes de cáncer de mama. Los atributos se corresponden con métricas calculadas sobre las células identificadas en imágenes de biopsias. Para cada paciente se realizan 10 métricas sobre varias células, y para cada métrica se registran la media, desviación estándar y peor resultado de todos valores. La clase a predecir es $0$ ó $1$ en función de que el tumor sea maligo o no."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Si quisieramos retornarlo ya en X e y lo hariamos así:\n",
    "(X,y) = load_breast_cancer(return_X_y=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['mean radius', 'mean texture', 'mean perimeter', 'mean area',\n",
      "       'mean smoothness', 'mean compactness', 'mean concavity',\n",
      "       'mean concave points', 'mean symmetry', 'mean fractal dimension',\n",
      "       'radius error', 'texture error', 'perimeter error', 'area error',\n",
      "       'smoothness error', 'compactness error', 'concavity error',\n",
      "       'concave points error', 'symmetry error', 'fractal dimension error',\n",
      "       'worst radius', 'worst texture', 'worst perimeter', 'worst area',\n",
      "       'worst smoothness', 'worst compactness', 'worst concavity',\n",
      "       'worst concave points', 'worst symmetry', 'worst fractal dimension',\n",
      "       'class'],\n",
      "      dtype='object')\n",
      "       mean radius  mean texture  mean perimeter    mean area  \\\n",
      "count   569.000000    569.000000      569.000000   569.000000   \n",
      "mean     14.127292     19.289649       91.969033   654.889104   \n",
      "std       3.524049      4.301036       24.298981   351.914129   \n",
      "min       6.981000      9.710000       43.790000   143.500000   \n",
      "25%      11.700000     16.170000       75.170000   420.300000   \n",
      "50%      13.370000     18.840000       86.240000   551.100000   \n",
      "75%      15.780000     21.800000      104.100000   782.700000   \n",
      "max      28.110000     39.280000      188.500000  2501.000000   \n",
      "\n",
      "       mean smoothness  mean compactness  mean concavity  mean concave points  \\\n",
      "count       569.000000        569.000000      569.000000           569.000000   \n",
      "mean          0.096360          0.104341        0.088799             0.048919   \n",
      "std           0.014064          0.052813        0.079720             0.038803   \n",
      "min           0.052630          0.019380        0.000000             0.000000   \n",
      "25%           0.086370          0.064920        0.029560             0.020310   \n",
      "50%           0.095870          0.092630        0.061540             0.033500   \n",
      "75%           0.105300          0.130400        0.130700             0.074000   \n",
      "max           0.163400          0.345400        0.426800             0.201200   \n",
      "\n",
      "       mean symmetry  mean fractal dimension     ...      worst texture  \\\n",
      "count     569.000000              569.000000     ...         569.000000   \n",
      "mean        0.181162                0.062798     ...          25.677223   \n",
      "std         0.027414                0.007060     ...           6.146258   \n",
      "min         0.106000                0.049960     ...          12.020000   \n",
      "25%         0.161900                0.057700     ...          21.080000   \n",
      "50%         0.179200                0.061540     ...          25.410000   \n",
      "75%         0.195700                0.066120     ...          29.720000   \n",
      "max         0.304000                0.097440     ...          49.540000   \n",
      "\n",
      "       worst perimeter   worst area  worst smoothness  worst compactness  \\\n",
      "count       569.000000   569.000000        569.000000         569.000000   \n",
      "mean        107.261213   880.583128          0.132369           0.254265   \n",
      "std          33.602542   569.356993          0.022832           0.157336   \n",
      "min          50.410000   185.200000          0.071170           0.027290   \n",
      "25%          84.110000   515.300000          0.116600           0.147200   \n",
      "50%          97.660000   686.500000          0.131300           0.211900   \n",
      "75%         125.400000  1084.000000          0.146000           0.339100   \n",
      "max         251.200000  4254.000000          0.222600           1.058000   \n",
      "\n",
      "       worst concavity  worst concave points  worst symmetry  \\\n",
      "count       569.000000            569.000000      569.000000   \n",
      "mean          0.272188              0.114606        0.290076   \n",
      "std           0.208624              0.065732        0.061867   \n",
      "min           0.000000              0.000000        0.156500   \n",
      "25%           0.114500              0.064930        0.250400   \n",
      "50%           0.226700              0.099930        0.282200   \n",
      "75%           0.382900              0.161400        0.317900   \n",
      "max           1.252000              0.291000        0.663800   \n",
      "\n",
      "       worst fractal dimension       class  \n",
      "count               569.000000  569.000000  \n",
      "mean                  0.083946    0.627417  \n",
      "std                   0.018061    0.483918  \n",
      "min                   0.055040    0.000000  \n",
      "25%                   0.071460    0.000000  \n",
      "50%                   0.080040    1.000000  \n",
      "75%                   0.092080    1.000000  \n",
      "max                   0.207500    1.000000  \n",
      "\n",
      "[8 rows x 31 columns]\n",
      "[0 1]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1d08f9f5ba8>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "# Acceder al dataset disponible en sklearn y crear el dataframe 'X' para los atributos, y la serie 'y' para la clase\n",
    "data_raw = load_breast_cancer()\n",
    "\n",
    "data = pd.DataFrame(data=data_raw['data'], columns=data_raw['feature_names'])\n",
    "data['class'] = data_raw['target']\n",
    "\n",
    "#    - Mostrar información sobre las columnas\n",
    "print(data.columns)\n",
    "\n",
    "#    - Mostrar indicadores estadísticos sobre las columnas\n",
    "print(data.describe())\n",
    "\n",
    "#    - Obtener una matriz numpy con los valores de los datos\n",
    "X = data.drop('class', axis=1).values\n",
    "y = data['class'].values\n",
    "\n",
    "#    - Mostrar los valores de la clase\n",
    "print(data['class'].unique())\n",
    "#    - Mostrar la distribución de los valores de la clase\n",
    "data['class'].plot(kind='hist')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Entrenamiento de un clasificador <a name=\"clasificador\"> </a>\n",
    "\n",
    "Entrenar un clasificador es muy simple en Sklearn. Basta con crear un objeto del estimador que queramos entrenar y ejecutar el método <code>fit</code>. En este notebook usaremos uno de los clasificadores que mejor resultado suele dar de los disponibles en Sklearn: <code>LogisticRegression</code>. \n",
    "\n",
    "### 1.1 ¿Qué es la regresión logística? ¿Es regresión o clasificación? <a name=\"logistica\"> </a>\n",
    "\n",
    "Es una técnica que se usa para predecir el resultado de una variable discreta, es por tanto una _técnica de clasificación_. En su formulación más simple (la binaria) se aprende a estimar la probabilidad de que una instancia pertenezca a la clase positiva de la siguiente forma:\n",
    "- Probabilidades cercanas a $1$ darán lugar a clasificar la instancia como de la clase positiva\n",
    "- Probabilidades cercanas a $0$ darán lugar a clasificar la instancia como de la clase negativa\n",
    "\n",
    "\n",
    "\n",
    "El modelo logístico establece la siguiente relación entre la probabilidad de pertenecer a la clase positiva y los valores de los atributos de la instancia $X$:\n",
    "\n",
    "$$\n",
    "P(X) = P(y=1\\;| \\;x_1, x_2, ..., x_n)  = \\frac{1}{1+exp(-\\alpha-\\beta_1x_1-\\beta_2x_2...-\\beta_nx_n)}\n",
    "$$\n",
    "\n",
    "A este tipo de funciones se les denomina _logísticas_ (de ahí el nombre _regresión logística_).\n",
    "\n",
    "Se puede generalizar a una clasificación no binaria (con $k$ categorías) mediante la construcción de $k-1$ clasificadores binarios.\n",
    "\n",
    "La función _logit_ nos permite convertir el problema de estimar la probablilidad en un problema de regresión lineal. Esta es la función _logit_:\n",
    "$$\n",
    "logit(p) = ln(\\frac{p}{1-p})\n",
    "$$\n",
    "\n",
    "Y mediante una serie de transformaciones se demuestra que:\n",
    "\n",
    "$$\n",
    "logit(P(y=1\\;| \\;x_1, x_2, ..., x_n)) = \\alpha+\\beta_1x_1+\\beta_2x_2...+\\beta_nx_n\n",
    "$$\n",
    "\n",
    "Gracias a usar la función _logit_ como función de enlace se consigue una formulación lineal del problema, lo que permite la aplicación de técnicas de regresión lineal (de ahí el nombre _regresión_ logística)  para el aprendizaje de los coeficientes $\\alpha$ y $\\beta_i$.\n",
    "\n",
    "\n",
    "\n",
    "### 1.2 Estimadores de Sklearn (_fit_, _predict_ y _predict proba_) <a name=\"estimadores\"> </a>\n",
    "\n",
    "<code>Sklearn</code> es un paquete muy complejo, y en este notebook solo usaremos algunos elementos del mismo. Los tres conceptos principales para comprender cómo se trabaja con <code>sklearn</code> son:\n",
    "- **Estimador**: es el mecanismo a través del cual se implementan los distintos algoritmos de aprendizaje automático disponibles en <code>sklearn</code>. Para crear un objeto estimador solo hace falta conocer el nombre de la clase que lo implementa. Por ejemplo <code>clasificador = LogisticRegression()</code>\n",
    "- **fit**: Es el método que se usa para entrenar un estimador. Es una propiedad del estimador, y basta con pasarle la matriz $X$ y el vector $y$. Por ejemplo, <code>clasificador.fit(X,y)</code> \n",
    "- **predict**: Es el método que se usa para predecir a partir de un estimador entrenada. Es una propiedad del estimador, y basta con pasarle la matriz $X\\_test$ con las instancias que se desean clasificar. Por ejemplo, <code>clasificador.predict(X_test)</code>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "          intercept_scaling=1, max_iter=100000, multi_class='warn',\n",
       "          n_jobs=None, penalty='l2', random_state=None, solver='lbfgs',\n",
       "          tol=0.0001, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: crear un estimador de la clase LogisticRegression y entrenarlo con el dataset <X,y>\n",
    "clf = LogisticRegression(solver='lbfgs', max_iter=100000)\n",
    "\n",
    "clf.fit(X, y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Una vez entrenado un clasificador, podemos usarlo para predecir la clase de un conjunto de instancias con el método <code>predict</code>. Muchos estimadores nos proporcionan también las probabilidades de cada una de las clases gracias al método <code>predict_proba</code>."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 0, 0, 1, 0, 0, 0, 0, 0, 0])"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: predecir la clase de los primeros 10 valores de X con el clasificador entrenado anteriormente\n",
    "clf.predict(X[0:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1.00000000e+00, 2.69187651e-15],\n",
       "       [9.99999980e-01, 2.00268142e-08],\n",
       "       [9.99999962e-01, 3.76756189e-08],\n",
       "       [4.93931576e-01, 5.06068424e-01],\n",
       "       [9.99991776e-01, 8.22423537e-06],\n",
       "       [8.00346768e-01, 1.99653232e-01],\n",
       "       [9.99999083e-01, 9.17266900e-07],\n",
       "       [9.82428673e-01, 1.75713269e-02],\n",
       "       [9.12940843e-01, 8.70591570e-02],\n",
       "       [9.97444109e-01, 2.55589139e-03]])"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: obtener la probabilidad de cada clase de los primeros 10 valores de X con el clasificador entrenado anteriormente\n",
    "clf.predict_proba(X[0:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds = clf.predict_proba(X[0:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1.00000000e+00, 2.69187651e-15],\n",
       "       [9.99999980e-01, 2.00268142e-08],\n",
       "       [9.99999962e-01, 3.76756189e-08],\n",
       "       [4.93931576e-01, 5.06068424e-01],\n",
       "       [9.99991776e-01, 8.22423537e-06],\n",
       "       [8.00346768e-01, 1.99653232e-01],\n",
       "       [9.99999083e-01, 9.17266900e-07],\n",
       "       [9.82428673e-01, 1.75713269e-02],\n",
       "       [9.12940843e-01, 8.70591570e-02],\n",
       "       [9.97444109e-01, 2.55589139e-03]])"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 0, 0, 1, 0, 0, 0, 0, 0, 0], dtype=int64)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Así se podrían convertir las probabilidades de nuevo en clases, si hiciera falta\n",
    "preds = np.argmax(preds,axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 0, 0, 1, 0, 0, 0, 0, 0, 0], dtype=int64)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Evaluación _train/test_ <a name=\"evaluacion_tt\"> </a>\n",
    "Las predicciones anteriores se han hecho sobre los mismos datos usados para entrenar el modelo. Para poder evaluar un clasificador de forma fiable es necesario usar datos distintos para el entrenamiento y la validación. Una manera de hacerlo es mediante una partición _train/test_ de los datos originales. En esta sección veremos cómo hacer este tipo de particiones con Sklearn y las utilizaremos para calcular una de las métricas de evaluación más comunes: el _accuracy_.\n",
    "\n",
    "### 2.1 Separación entrenamiento y validación <a name=\"separacion\"> </a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: dividir el corpus <X, y> en dos <X_train, y_train> y <X_test, y_test> con una distribución 80%-20%\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20, random_state=123)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "          intercept_scaling=1, max_iter=100000, multi_class='warn',\n",
       "          n_jobs=None, penalty='l2', random_state=None, solver='lbfgs',\n",
       "          tol=0.0001, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: entrenar un clasificador con el corpus <X_train, y_train>\n",
    "clf = LogisticRegression(solver='lbfgs', max_iter=100000)\n",
    "\n",
    "clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1, 0, 0,\n",
       "       1, 1, 1, 0, 0, 1, 0, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0, 1,\n",
       "       0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 0, 1, 1, 1, 1, 0,\n",
       "       1, 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0, 1,\n",
       "       0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "       1, 1, 1, 0])"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: predecir la clase para las instancias de X_test y guardar las predicciones en y_test_pred\n",
    "y_test_pred = clf.predict(X_test)\n",
    "y_test_pred"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.2 Cálculo del accuracy <a name=\"accuracy\"> </a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9824561403508771"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: calcular el accuracy de las predicciones comparando y_test e y_test_pred\n",
    "(y_test == y_test_pred).sum()/(y_test.shape[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.3 Probabilidades de pertenencia <a name=\"pertenencia\"> </a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: calcular las probabilidades de cada clase para las instancias de X_test y guardarlas en y_test_proba\n",
    "y_test_proba = clf.predict_proba(X_test)\n",
    "y_test_proba"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Como podemos observar aparecen dos probabilidades para cada fila: la probabilidad de elegir la clase $0$ y la probabilidad de elegir la clase $1$. Un análisis de la distribución de estas probabilidades, nos dará pistas sobre si el conjunto de datos a clasificar está _bien separado_. Si es así, habrá pocos casos dudosos al clasificador le costará menos trabajo decidir."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0, 1)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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: mostrar con un histograma la distribución de la probabilidad de pertenecer a la clase 1\n",
    "sns.distplot(y_test_proba[:, 1], bins=15)\n",
    "plt.xlim(0, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0, 1)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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: mostrar con un histograma la distribución de la probabilidad de pertenecer a la clase 0\n",
    "sns.distplot(y_test_proba[:, 0], bins=15)\n",
    "plt.xlim(0, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Evaluación mediante validación cruzada <a name=\"evaluacion_cruzada\"> </a>\n",
    "\n",
    "La evaluación _train/test_ tiene dos principales inconvenientes:\n",
    "- Desperdicia parte del material de entrenamiento.\n",
    "- Los resultados son dependientes de la partición, pudiendo ser sensiblemente distintos con otras particiones.\n",
    "\n",
    "La validación cruzada es un medio para mitigar ambos efectos. Consiste en dividir el dataset en varias carpetas y realizar diferentes evaluaciones usando en cada una de ellas una carpeta distinta como _test_ y el resto de carpetas como _train_. El resultado final de la evaluación se puede calcular mediante la media de todas las evaluaciones.\n",
    "\n",
    "Los métodos que usaremos son:\n",
    "- **cross_val_predict**: usa por turnos todas las instancias de $X$ tanto para entrenar, como para predecir, y devuelve un vector $y\\_pred$ con las predicciones obtenidas para todas las instancias de $x$. \n",
    "- **cross_val_score**: usa por turnos todas las instancias de $X$ tanto para entrenar, como para evaluar, y devuelve una métrica de evaluación (por defecto es la proporción de aciertos)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: predecir la clase de todas las instancias mediante validación cruzada y guardar las prediccciones en y_pred\n",
    "clf = LogisticRegression(solver='lbfgs', max_iter=100000)\n",
    "\n",
    "y_pred = cross_val_predict(clf, X, y, cv=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: calcular las probabilidades de cada clase para todas las instancias mediante validación cruzada y guardarlas en y_proba\n",
    "clf = LogisticRegression(solver='lbfgs', max_iter=100000)\n",
    "\n",
    "y_proba = cross_val_predict(clf, X, y, cv=5, method='predict_proba')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: calcular el accuracy de las predicciones sobre todas las instancias mediante validación cruzada\n",
    "clf = LogisticRegression(solver='lbfgs', max_iter=100000)\n",
    "\n",
    "# retornara el accuracy para cada uno de los folds.\n",
    "y_acc = cross_val_score(clf, X, y, cv=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.93913043, 0.94782609, 0.98230088, 0.92920354, 0.95575221])"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Belladona\\Anaconda3\\envs\\ztdl\\lib\\site-packages\\scipy\\stats\\stats.py:1713: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.\n",
      "  return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1d0b3220fd0>"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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:\n",
    "#    - Calcular el accuracy con distintos números de carpetas en la validación cruzada (p.e. de 2 a 10)\n",
    "#    - Mostrar los resultados en una gráfica\n",
    "\n",
    "scores = list()\n",
    "folds = list(range(2, 11))\n",
    "\n",
    "for fold in folds:\n",
    "    clf = LogisticRegression(solver='lbfgs', max_iter=100000)\n",
    "\n",
    "    y_proba = cross_val_score(clf, X, y, cv=fold).mean()\n",
    "    \n",
    "    scores.append(y_proba)\n",
    "\n",
    "sns.regplot(x='x', y='y', data=pd.DataFrame({'x': folds, 'y': scores}))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Matriz de confusión, _accuracy_, precisón, cobertura, medida f1, TPR y FPR <a name=\"matriz_cf\"> </a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A partir de esta sección trabajaremos sobre los resultados de la validación cruzada (mediante las series <code>y</code>, <code>y_pred</code> e <code>y_proba</code>) aunque el análisis podía haberse hecho perfectamente con una evaluación _train/test_. Mostraremos cómo calcular distintas métricas de evaluación. En muchas ocasiones lo haremos de dos formas distintas:\n",
    "- Con un método ofrecido por Sklearn que suele recibir <code>y</code> e <code>y_pred</code>.\n",
    "- A partir de las celdas de la matriz de confusión.\n",
    "\n",
    "### 4.1 Número de aciertos y fallos: TP, FP, TN, FN <a name=\"aciertos_fallos\"> </a>\n",
    "\n",
    "La matriz de confusión muestra el número de veces que se han producido los distintos tipos de aciertos y fallos. En una clasificación binaria tiene cuatro celdas que suelen denominarse con los siguientes nombres:\n",
    "- TP: _true positives_\n",
    "- TN: _true negatives_\n",
    "- FP: _false positives_\n",
    "- FN: _false negatives_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = LogisticRegression(solver='lbfgs', max_iter=100000)\n",
    "\n",
    "y_pred = cross_val_predict(clf, X, y, cv=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = LogisticRegression(solver='lbfgs', max_iter=100000)\n",
    "\n",
    "y_proba = cross_val_predict(clf, X, y, cv=5, method='predict_proba')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1d0aed542e8>"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "# EJERCICIO: crear la matriz de confusión y mostrarla mediante un mapa de calor\n",
    "sns.heatmap(metrics.confusion_matrix(y, y_pred),annot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: guardar los valores de la matriz de confusión en sus variables correspondientes (TP, TN, FP, FN)\n",
    "TN, FP, FN, TP = metrics.confusion_matrix(y, y_pred).ravel() # ravel() returns a flattened array"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Una de las métricas más usadas es el _accuracy_ que mide directamente el porcentaje de aciertos. Si el dataset está balanceado es un buen indicador, pero si alguna de las clases es muy mayoritaria (o minoritaria) la información que nos dá la métrica puede ser bastante engañosa."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9507908611599297\n",
      "0.9507908611599297\n"
     ]
    }
   ],
   "source": [
    "# EJERCICIO: calcular la medida 'accuracy' de dos formas distintas:\n",
    "#    - Con el método proporcionado por Sklearn\n",
    "#    - A partir de los componentes de la matriz de confusión\n",
    "# accuracy = (TP + TN) / (TP + TN + FP + FN)\n",
    "\n",
    "print(metrics.accuracy_score(y, y_pred))\n",
    "\n",
    "accuracy = (TP + TN) / (TP + TN + FP + FN)\n",
    "\n",
    "print(accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: definir una función que reciba un clasificador y muestre el accuracy obtenido y la matriz de confusión\n",
    "def get_metrics(classifier, X, y):\n",
    "    y_pred = cross_val_predict(clf, X, y, cv=5)\n",
    "    print(f'Accuracy: {metrics.accuracy_score(y, y_pred)*100}%')\n",
    "    print(f'Confusion matrix: {metrics.confusion_matrix(y, y_pred)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.2 accuracy, precision, recall y f1 <a name=\"medidas\"> </a>\n",
    "\n",
    "El siguiente paquete de métricas es el formado por la precisión, cobertura (_recall_) y medida f1. Son métricas que dan diferente importancia al tipo de error (p.e. falsos positivos o falsos negativos en clasificación binaria). Pueden ser de utilidad para sistemas en los que nos preocupan más unos errores que otros: por ejemplo, es menos grave dejar pasar un correo _spam_ que eliminar un correo correcto. El significado intuitivo de las métricas es el siguiente:\n",
    "- _precision_: grado de acierto en las instancias propuestas como positivas.\n",
    "- _recall_: porcentaje de recuperación del total de las instancias positivas.\n",
    "- _f1_ : media armónica de precisión y cobertura.\n",
    "\n",
    "Al combinar dos métricas complementarias, la medida _f1_ es apropiada para datasets que no estén bien balanceados."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9532967032967034\n",
      "0.9532967032967034\n"
     ]
    }
   ],
   "source": [
    "# EJERCICIO: calcular la medida 'precisión' de dos formas distintas:\n",
    "#    - Con el método proporcionado por Sklearn\n",
    "#    - A partir de los componentes de la matriz de confusión\n",
    "print(metrics.precision_score(y, y_pred))\n",
    "print(TP/(TP+FP))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9719887955182073\n",
      "0.9719887955182073\n"
     ]
    }
   ],
   "source": [
    "# EJERCICIO: calcular la medida 'cobertura' (o recall) de dos formas distintas:\n",
    "#    - Con el método proporcionado por Sklearn\n",
    "#    - A partir de los componentes de la matriz de confusión\n",
    "print(metrics.recall_score(y, y_pred))\n",
    "print(TP/(TP+FN))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9625520110957004\n",
      "0.9625520110957004\n"
     ]
    }
   ],
   "source": [
    "# EJERCICIO: calcular la medida 'f1' de dos formas distintas:\n",
    "#    - Con el método proporcionado por Sklearn\n",
    "#    - A partir de las medidas 'precisión' y 'cobertura'\n",
    "print(metrics.f1_score(y, y_pred))\n",
    "print((2*TP)/((2*TP) + FP + FN))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.3 _True Positive Rate_ y _False Positive Rate_ <a name=\"rates\"> </a>\n",
    "Las siguientes dos métricas también suelen ir de la mano y, al igual que la precisión y la cobertura, son complementarias. Se trata de TPR (_True Positive Rate_) y FPR (_False Positive Rate_) y se definen de la siguiente forma:\n",
    "\n",
    "- TPR: positivos correctos (TP) dividido entre todos los positivos (TP+FN). También se denomina a esta métrica _sensibilidad_ y _cobertura_.\n",
    "- FPR: positivos erróneos (FP) dividido entre todos los negativos (FP+TN). A la métrica $1-FPR$ también se le denomina _especificidad_.\n",
    "\n",
    "Ambas métricas se suelen combinar en una curva, denominada ROC, que veremos en la siguiente sección."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9719887955182073\n"
     ]
    }
   ],
   "source": [
    "# EJERCICIO: calcular la medida TPR a partir de los componentes de la matriz de confusión\n",
    "TPR = TP / (TP+FN)\n",
    "print(TPR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.08018867924528301\n"
     ]
    }
   ],
   "source": [
    "# EJERCICIO: calcular la medida FPR a partir de los componentes de la matriz de confusión\n",
    "FPR = FP / (FP+TN)\n",
    "print(FPR)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. Curva ROC (TPR-FPR) y curva _Precision-Recall_ <a name=\"curvas\"> </a>\n",
    "\n",
    "Ambas curvas (y sus correspondientes métricas de agregación) se benefician de la complementariedad de las métricas en las que se apoyan para definir unas métricas que tengan en cuenta los distintos tipos de errores que se pueden cometer.\n",
    "\n",
    "La principal contribución de estas curvas con respecto a otras métricas es que muestran la calidad del clasificador en distintos escenarios de decisión, diferenciados por el umbral de probabilidad usado para determinar la clase. Esto se puede hacer siempre que dispongamos las probabilidades generadas por <code>predict_proba</code>:\n",
    "\n",
    "![image.png](attachment:image.png)\n",
    "\n",
    "Probando distintos umbrales de decisión, obtendremos métricas distintas. Y es, precisamente, con esos puntos con los que se trazan ambas curvas. La definición intuitiva de cada una de ellas es la siguente:\n",
    "- Curva ROC: unión de puntos (FPR, TPR) con distintos umbrales de decisión. La métrica ROC-AUC resume la curva en un único valor numérico mediante el área que queda bajo ella.\n",
    "- Curva _Precision-Recall_: unión de puntos (_recall_, _precision_) con distintos umbrales de decisión. La métrica _average precision score_ resume la curva en un único valor numérico mediante el área que queda bajo ella.\n",
    "\n",
    "El hecho de que ambas curvas integren métricas complementarias y que se apoyen en distintos umbrales de decisión, las hacen muy apropiadas para situaciones en las que el dataset no está balanceado."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Antes de nada debemos disponer de las probabilidades para la clase 1\n",
    "# Usaremos 'y_proba' (validación cruzada), aunque también podría hacerse sobre 'y_test_proba' (train/test)\n",
    "y_proba_clase1 =  pd.Series(y_proba[:,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([0.        , 0.        , 0.        , 0.00471698, 0.00471698,\n",
       "        0.00943396, 0.00943396, 0.01886792, 0.01886792, 0.02358491,\n",
       "        0.02358491, 0.02830189, 0.02830189, 0.03301887, 0.03301887,\n",
       "        0.04245283, 0.04245283, 0.04716981, 0.04716981, 0.05188679,\n",
       "        0.05188679, 0.05660377, 0.05660377, 0.07075472, 0.07075472,\n",
       "        0.0754717 , 0.0754717 , 0.08018868, 0.08018868, 0.08490566,\n",
       "        0.08490566, 0.10377358, 0.10377358, 0.13207547, 0.13207547,\n",
       "        0.13679245, 0.13679245, 0.14622642, 0.14622642, 0.16509434,\n",
       "        0.16509434, 1.        ]),\n",
       " array([0.        , 0.00280112, 0.59943978, 0.59943978, 0.77871148,\n",
       "        0.77871148, 0.87955182, 0.87955182, 0.90196078, 0.90196078,\n",
       "        0.9047619 , 0.9047619 , 0.93277311, 0.93277311, 0.93557423,\n",
       "        0.93557423, 0.93837535, 0.93837535, 0.94397759, 0.94397759,\n",
       "        0.95798319, 0.95798319, 0.96358543, 0.96358543, 0.96638655,\n",
       "        0.96638655, 0.96918768, 0.96918768, 0.9719888 , 0.9719888 ,\n",
       "        0.97478992, 0.97478992, 0.97759104, 0.97759104, 0.98319328,\n",
       "        0.98319328, 0.98879552, 0.98879552, 0.99159664, 0.99159664,\n",
       "        1.        , 1.        ]),\n",
       " array([1.99999996e+00, 9.99999956e-01, 9.95127996e-01, 9.95127070e-01,\n",
       "        9.73405407e-01, 9.72224801e-01, 9.22462531e-01, 9.17701902e-01,\n",
       "        8.74577190e-01, 8.68181096e-01, 8.64042662e-01, 8.61979507e-01,\n",
       "        8.24370214e-01, 8.15521192e-01, 8.12942211e-01, 7.95991775e-01,\n",
       "        7.94051738e-01, 7.90888333e-01, 7.61330528e-01, 7.21934037e-01,\n",
       "        6.93931259e-01, 6.62021611e-01, 6.51997657e-01, 6.26029954e-01,\n",
       "        5.64035347e-01, 5.61205932e-01, 5.46813510e-01, 5.08591922e-01,\n",
       "        4.99001892e-01, 4.96316036e-01, 4.51602100e-01, 3.55908866e-01,\n",
       "        3.53307630e-01, 2.53945835e-01, 1.96556361e-01, 1.94823262e-01,\n",
       "        1.83102767e-01, 1.68856871e-01, 1.44871668e-01, 1.23422640e-01,\n",
       "        9.07293958e-02, 2.00904158e-38]))"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: calcular los puntos de la curva ROC mediante el método proporcionado por Sklearn\n",
    "metrics.roc_curve(y, y_proba_clase1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "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: mostrar gráficamente la curva ROC\n",
    "skplt.metrics.plot_roc(y, y_proba)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9907774430526928"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: calcular la métrica ROC-AUC mediante el método proporcionado por Sklearn\n",
    "metrics.roc_auc_score(y, y_proba_clase1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: definir una función que reciba un clasificador y muestre la siguiente información:\n",
    "#       - valores FPRs y TPRs\n",
    "#       - valores de los umbrales para la curva roc\n",
    "#       - valor del área bajo la curva\n",
    "#       - gráfica del área bajo la curva\n",
    "#\n",
    "\n",
    "def get_metrics2(classifier, X, y):\n",
    "    y_pred = cross_val_predict(clf, X, y, cv=5)\n",
    "    y_proba = cross_val_predict(clf, X, y, cv=5, method='predict_proba')\n",
    "    \n",
    "    TN, FP, FN, TP = metrics.confusion_matrix(y, y_pred).ravel()\n",
    "    \n",
    "    TPR = TP / (TP+FN)\n",
    "    print(f'TPR: {TPR}')\n",
    "    \n",
    "    FPR = FP / (FP+TN)\n",
    "    print(f'FPR: {FPR}')\n",
    "    \n",
    "    y_proba_clase1 =  pd.Series(y_proba[:,1])\n",
    "    print(f'ROC Curve: {metrics.roc_curve(y, y_proba_clase1)}')\n",
    "    \n",
    "    print(f'AUC ROC: {metrics.roc_auc_score(y, y_proba_clase1)}')\n",
    "    \n",
    "    skplt.metrics.plot_roc(y, y_proba)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([0.91071429, 0.91048593, 0.91025641, 0.91002571, 0.91237113,\n",
       "        0.9121447 , 0.91450777, 0.91688312, 0.91927083, 0.91906005,\n",
       "        0.92146597, 0.92388451, 0.92368421, 0.92612137, 0.92592593,\n",
       "        0.92572944, 0.92819149, 0.928     , 0.93048128, 0.93297587,\n",
       "        0.93548387, 0.93800539, 0.94054054, 0.94308943, 0.94565217,\n",
       "        0.94822888, 0.94808743, 0.95068493, 0.9532967 , 0.95592287,\n",
       "        0.9558011 , 0.95567867, 0.95833333, 0.96100279, 0.96368715,\n",
       "        0.96358543, 0.96629213, 0.96619718, 0.96610169, 0.96600567,\n",
       "        0.96590909, 0.96581197, 0.96857143, 0.96848138, 0.9683908 ,\n",
       "        0.97118156, 0.97109827, 0.97101449, 0.97383721, 0.97667638,\n",
       "        0.97953216, 0.98240469, 0.98235294, 0.98230088, 0.98224852,\n",
       "        0.98219585, 0.98214286, 0.98208955, 0.98203593, 0.98198198,\n",
       "        0.98192771, 0.98489426, 0.98484848, 0.98480243, 0.9847561 ,\n",
       "        0.98776758, 0.98773006, 0.98769231, 0.98765432, 0.9876161 ,\n",
       "        0.98757764, 0.98753894, 0.9875    , 0.98746082, 0.98742138,\n",
       "        0.9873817 , 0.99050633, 0.99365079, 0.99363057, 0.99361022,\n",
       "        0.99358974, 0.99356913, 0.99354839, 0.99352751, 0.99350649,\n",
       "        0.99348534, 0.99346405, 0.99344262, 0.99342105, 0.99339934,\n",
       "        0.99337748, 0.99335548, 0.99333333, 0.99331104, 0.99328859,\n",
       "        0.99326599, 0.99324324, 0.99322034, 0.99319728, 0.99317406,\n",
       "        0.99315068, 0.99312715, 0.99310345, 0.99307958, 0.99305556,\n",
       "        0.99303136, 0.99300699, 0.99298246, 0.99295775, 0.99293286,\n",
       "        0.9929078 , 0.99288256, 0.99642857, 0.99641577, 0.99640288,\n",
       "        0.99638989, 0.99637681, 0.99636364, 0.99635036, 0.996337  ,\n",
       "        0.99632353, 0.99630996, 0.9962963 , 0.99628253, 0.99626866,\n",
       "        0.99625468, 0.9962406 , 0.99622642, 0.99621212, 0.99619772,\n",
       "        0.99618321, 0.99616858, 0.99615385, 0.996139  , 0.99612403,\n",
       "        0.99610895, 0.99609375, 0.99607843, 0.99606299, 0.99604743,\n",
       "        0.99603175, 0.99601594, 0.996     , 0.99598394, 0.99596774,\n",
       "        0.99595142, 0.99593496, 0.99591837, 0.99590164, 0.99588477,\n",
       "        0.99586777, 0.99585062, 0.99583333, 0.9958159 , 0.99579832,\n",
       "        0.99578059, 0.99576271, 0.99574468, 0.9957265 , 0.99570815,\n",
       "        0.99568966, 0.995671  , 0.99565217, 0.99563319, 0.99561404,\n",
       "        0.99559471, 0.99557522, 0.99555556, 0.99553571, 0.9955157 ,\n",
       "        0.9954955 , 0.99547511, 0.99545455, 0.99543379, 0.99541284,\n",
       "        0.99539171, 0.99537037, 0.99534884, 0.9953271 , 0.99530516,\n",
       "        0.99528302, 0.99526066, 0.9952381 , 0.99521531, 0.99519231,\n",
       "        0.99516908, 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        , 1.        , 1.        ,\n",
       "        1.        , 1.        , 1.        ]),\n",
       " array([1.        , 0.99719888, 0.99439776, 0.99159664, 0.99159664,\n",
       "        0.98879552, 0.98879552, 0.98879552, 0.98879552, 0.9859944 ,\n",
       "        0.9859944 , 0.9859944 , 0.98319328, 0.98319328, 0.98039216,\n",
       "        0.97759104, 0.97759104, 0.97478992, 0.97478992, 0.97478992,\n",
       "        0.97478992, 0.97478992, 0.97478992, 0.97478992, 0.97478992,\n",
       "        0.97478992, 0.9719888 , 0.9719888 , 0.9719888 , 0.9719888 ,\n",
       "        0.96918768, 0.96638655, 0.96638655, 0.96638655, 0.96638655,\n",
       "        0.96358543, 0.96358543, 0.96078431, 0.95798319, 0.95518207,\n",
       "        0.95238095, 0.94957983, 0.94957983, 0.94677871, 0.94397759,\n",
       "        0.94397759, 0.94117647, 0.93837535, 0.93837535, 0.93837535,\n",
       "        0.93837535, 0.93837535, 0.93557423, 0.93277311, 0.92997199,\n",
       "        0.92717087, 0.92436975, 0.92156863, 0.91876751, 0.91596639,\n",
       "        0.91316527, 0.91316527, 0.91036415, 0.90756303, 0.9047619 ,\n",
       "        0.9047619 , 0.90196078, 0.89915966, 0.89635854, 0.89355742,\n",
       "        0.8907563 , 0.88795518, 0.88515406, 0.88235294, 0.87955182,\n",
       "        0.8767507 , 0.8767507 , 0.8767507 , 0.87394958, 0.87114846,\n",
       "        0.86834734, 0.86554622, 0.8627451 , 0.85994398, 0.85714286,\n",
       "        0.85434174, 0.85154062, 0.8487395 , 0.84593838, 0.84313725,\n",
       "        0.84033613, 0.83753501, 0.83473389, 0.83193277, 0.82913165,\n",
       "        0.82633053, 0.82352941, 0.82072829, 0.81792717, 0.81512605,\n",
       "        0.81232493, 0.80952381, 0.80672269, 0.80392157, 0.80112045,\n",
       "        0.79831933, 0.79551821, 0.79271709, 0.78991597, 0.78711485,\n",
       "        0.78431373, 0.78151261, 0.78151261, 0.77871148, 0.77591036,\n",
       "        0.77310924, 0.77030812, 0.767507  , 0.76470588, 0.76190476,\n",
       "        0.75910364, 0.75630252, 0.7535014 , 0.75070028, 0.74789916,\n",
       "        0.74509804, 0.74229692, 0.7394958 , 0.73669468, 0.73389356,\n",
       "        0.73109244, 0.72829132, 0.7254902 , 0.72268908, 0.71988796,\n",
       "        0.71708683, 0.71428571, 0.71148459, 0.70868347, 0.70588235,\n",
       "        0.70308123, 0.70028011, 0.69747899, 0.69467787, 0.69187675,\n",
       "        0.68907563, 0.68627451, 0.68347339, 0.68067227, 0.67787115,\n",
       "        0.67507003, 0.67226891, 0.66946779, 0.66666667, 0.66386555,\n",
       "        0.66106443, 0.65826331, 0.65546218, 0.65266106, 0.64985994,\n",
       "        0.64705882, 0.6442577 , 0.64145658, 0.63865546, 0.63585434,\n",
       "        0.63305322, 0.6302521 , 0.62745098, 0.62464986, 0.62184874,\n",
       "        0.61904762, 0.6162465 , 0.61344538, 0.61064426, 0.60784314,\n",
       "        0.60504202, 0.6022409 , 0.59943978, 0.59663866, 0.59383754,\n",
       "        0.59103641, 0.58823529, 0.58543417, 0.58263305, 0.57983193,\n",
       "        0.57703081, 0.57703081, 0.57422969, 0.57142857, 0.56862745,\n",
       "        0.56582633, 0.56302521, 0.56022409, 0.55742297, 0.55462185,\n",
       "        0.55182073, 0.54901961, 0.54621849, 0.54341737, 0.54061625,\n",
       "        0.53781513, 0.53501401, 0.53221289, 0.52941176, 0.52661064,\n",
       "        0.52380952, 0.5210084 , 0.51820728, 0.51540616, 0.51260504,\n",
       "        0.50980392, 0.5070028 , 0.50420168, 0.50140056, 0.49859944,\n",
       "        0.49579832, 0.4929972 , 0.49019608, 0.48739496, 0.48459384,\n",
       "        0.48179272, 0.4789916 , 0.47619048, 0.47338936, 0.47058824,\n",
       "        0.46778711, 0.46498599, 0.46218487, 0.45938375, 0.45658263,\n",
       "        0.45378151, 0.45098039, 0.44817927, 0.44537815, 0.44257703,\n",
       "        0.43977591, 0.43697479, 0.43417367, 0.43137255, 0.42857143,\n",
       "        0.42577031, 0.42296919, 0.42016807, 0.41736695, 0.41456583,\n",
       "        0.41176471, 0.40896359, 0.40616246, 0.40336134, 0.40056022,\n",
       "        0.3977591 , 0.39495798, 0.39215686, 0.38935574, 0.38655462,\n",
       "        0.3837535 , 0.38095238, 0.37815126, 0.37535014, 0.37254902,\n",
       "        0.3697479 , 0.36694678, 0.36414566, 0.36134454, 0.35854342,\n",
       "        0.3557423 , 0.35294118, 0.35014006, 0.34733894, 0.34453782,\n",
       "        0.34173669, 0.33893557, 0.33613445, 0.33333333, 0.33053221,\n",
       "        0.32773109, 0.32492997, 0.32212885, 0.31932773, 0.31652661,\n",
       "        0.31372549, 0.31092437, 0.30812325, 0.30532213, 0.30252101,\n",
       "        0.29971989, 0.29691877, 0.29411765, 0.29131653, 0.28851541,\n",
       "        0.28571429, 0.28291317, 0.28011204, 0.27731092, 0.2745098 ,\n",
       "        0.27170868, 0.26890756, 0.26610644, 0.26330532, 0.2605042 ,\n",
       "        0.25770308, 0.25490196, 0.25210084, 0.24929972, 0.2464986 ,\n",
       "        0.24369748, 0.24089636, 0.23809524, 0.23529412, 0.232493  ,\n",
       "        0.22969188, 0.22689076, 0.22408964, 0.22128852, 0.21848739,\n",
       "        0.21568627, 0.21288515, 0.21008403, 0.20728291, 0.20448179,\n",
       "        0.20168067, 0.19887955, 0.19607843, 0.19327731, 0.19047619,\n",
       "        0.18767507, 0.18487395, 0.18207283, 0.17927171, 0.17647059,\n",
       "        0.17366947, 0.17086835, 0.16806723, 0.16526611, 0.16246499,\n",
       "        0.15966387, 0.15686275, 0.15406162, 0.1512605 , 0.14845938,\n",
       "        0.14565826, 0.14285714, 0.14005602, 0.1372549 , 0.13445378,\n",
       "        0.13165266, 0.12885154, 0.12605042, 0.1232493 , 0.12044818,\n",
       "        0.11764706, 0.11484594, 0.11204482, 0.1092437 , 0.10644258,\n",
       "        0.10364146, 0.10084034, 0.09803922, 0.0952381 , 0.09243697,\n",
       "        0.08963585, 0.08683473, 0.08403361, 0.08123249, 0.07843137,\n",
       "        0.07563025, 0.07282913, 0.07002801, 0.06722689, 0.06442577,\n",
       "        0.06162465, 0.05882353, 0.05602241, 0.05322129, 0.05042017,\n",
       "        0.04761905, 0.04481793, 0.04201681, 0.03921569, 0.03641457,\n",
       "        0.03361345, 0.03081232, 0.0280112 , 0.02521008, 0.02240896,\n",
       "        0.01960784, 0.01680672, 0.0140056 , 0.01120448, 0.00840336,\n",
       "        0.00560224, 0.00280112, 0.        ]),\n",
       " array([0.09756471, 0.10852899, 0.11273351, 0.12525037, 0.12573998,\n",
       "        0.12949164, 0.12981883, 0.14092223, 0.14943048, 0.16326033,\n",
       "        0.1703543 , 0.19482096, 0.19949095, 0.20410213, 0.23290038,\n",
       "        0.25308165, 0.2685761 , 0.27919084, 0.28023839, 0.29278598,\n",
       "        0.33078881, 0.34904275, 0.37732521, 0.40597598, 0.44341948,\n",
       "        0.45677637, 0.46237457, 0.49163485, 0.50857391, 0.52604873,\n",
       "        0.54470752, 0.55153758, 0.62368813, 0.63672968, 0.63958691,\n",
       "        0.64111761, 0.65026153, 0.65770132, 0.68853968, 0.70047998,\n",
       "        0.71969815, 0.72704417, 0.76129319, 0.77270682, 0.78841362,\n",
       "        0.78960251, 0.79043918, 0.79542778, 0.80012227, 0.81162314,\n",
       "        0.81355375, 0.8163751 , 0.81724143, 0.82363198, 0.83204176,\n",
       "        0.83331459, 0.84058852, 0.84582142, 0.85384267, 0.85426157,\n",
       "        0.85509593, 0.85729107, 0.86103579, 0.86450474, 0.86665526,\n",
       "        0.86977345, 0.87937256, 0.88085116, 0.88517077, 0.90026975,\n",
       "        0.90798692, 0.9110442 , 0.9125701 , 0.91305604, 0.91449222,\n",
       "        0.91604409, 0.91688294, 0.921737  , 0.92319843, 0.92417935,\n",
       "        0.93569838, 0.93840002, 0.93880818, 0.93896915, 0.94074548,\n",
       "        0.94246169, 0.94271505, 0.94308321, 0.94350707, 0.94729668,\n",
       "        0.94850175, 0.94897995, 0.95069537, 0.95076292, 0.950887  ,\n",
       "        0.95208575, 0.95561661, 0.95581398, 0.95690972, 0.95772783,\n",
       "        0.95781231, 0.95951891, 0.95986575, 0.96051885, 0.96071095,\n",
       "        0.96074134, 0.96136277, 0.9627751 , 0.96668687, 0.96698645,\n",
       "        0.96769032, 0.96865517, 0.96951854, 0.97028338, 0.9716965 ,\n",
       "        0.97190122, 0.97293107, 0.97459104, 0.9758667 , 0.9764557 ,\n",
       "        0.97670027, 0.97727975, 0.97748826, 0.97775076, 0.97781147,\n",
       "        0.978291  , 0.97917937, 0.97951658, 0.98021174, 0.98047839,\n",
       "        0.98086835, 0.98089557, 0.98118451, 0.98134089, 0.98146001,\n",
       "        0.98199553, 0.98213347, 0.98245699, 0.98289349, 0.98314709,\n",
       "        0.98316452, 0.98340168, 0.98343329, 0.98359212, 0.98484264,\n",
       "        0.98504006, 0.98562618, 0.98567863, 0.98603277, 0.98634381,\n",
       "        0.98716821, 0.9874087 , 0.98746499, 0.9875116 , 0.98755753,\n",
       "        0.98841512, 0.98856633, 0.98870334, 0.99040082, 0.9905494 ,\n",
       "        0.99065956, 0.9913911 , 0.99146909, 0.99161088, 0.99178029,\n",
       "        0.99181028, 0.99197664, 0.99224368, 0.99257827, 0.99283127,\n",
       "        0.99355971, 0.99364234, 0.99368046, 0.99388252, 0.99395703,\n",
       "        0.99401494, 0.99429187, 0.99433663, 0.99435544, 0.99451372,\n",
       "        0.99468791, 0.99479656, 0.99491168, 0.99494347, 0.99500635,\n",
       "        0.99518254, 0.99528423, 0.99532211, 0.99532593, 0.99550915,\n",
       "        0.99551939, 0.99552221, 0.99572208, 0.99575602, 0.99585205,\n",
       "        0.99596231, 0.99610262, 0.99612823, 0.9962719 , 0.99631786,\n",
       "        0.99637122, 0.99660582, 0.99682174, 0.99683121, 0.99704776,\n",
       "        0.99709652, 0.9971137 , 0.99711519, 0.99712588, 0.99723584,\n",
       "        0.99732672, 0.99750215, 0.99752556, 0.99754721, 0.99757477,\n",
       "        0.99773476, 0.99776583, 0.99778296, 0.99780107, 0.99783881,\n",
       "        0.99784256, 0.99790354, 0.99794802, 0.99795325, 0.9979796 ,\n",
       "        0.99799623, 0.99800867, 0.99811799, 0.99812923, 0.9981773 ,\n",
       "        0.99819495, 0.99822416, 0.99823908, 0.99824894, 0.99829461,\n",
       "        0.99830061, 0.99830192, 0.99831537, 0.99831686, 0.99832546,\n",
       "        0.99833825, 0.99834844, 0.99842972, 0.99851544, 0.99857152,\n",
       "        0.99861383, 0.99867222, 0.99868447, 0.99877485, 0.99882047,\n",
       "        0.99882356, 0.99884796, 0.998962  , 0.99898355, 0.99900589,\n",
       "        0.99902617, 0.99903929, 0.99904745, 0.99906433, 0.99906853,\n",
       "        0.99907102, 0.9991121 , 0.99912806, 0.99913175, 0.99913689,\n",
       "        0.99913692, 0.99914483, 0.99916063, 0.99916755, 0.99920121,\n",
       "        0.99921   , 0.99922322, 0.99922993, 0.99925857, 0.99928604,\n",
       "        0.99928826, 0.99932377, 0.99936222, 0.99936474, 0.9993987 ,\n",
       "        0.99944207, 0.99947049, 0.99947548, 0.99948255, 0.9994885 ,\n",
       "        0.99949011, 0.9994976 , 0.99949973, 0.99951501, 0.99951712,\n",
       "        0.99953733, 0.99955271, 0.99957479, 0.9995805 , 0.99958219,\n",
       "        0.99958363, 0.99958751, 0.99960136, 0.99960173, 0.999606  ,\n",
       "        0.99961061, 0.99961944, 0.99962957, 0.99963171, 0.99965082,\n",
       "        0.99965195, 0.99965418, 0.99965538, 0.99965722, 0.99965727,\n",
       "        0.99966386, 0.99966806, 0.99967076, 0.99967097, 0.99970535,\n",
       "        0.99971374, 0.99971409, 0.99972729, 0.99974361, 0.99974381,\n",
       "        0.99976   , 0.9997644 , 0.99977312, 0.99979392, 0.99979631,\n",
       "        0.9997968 , 0.99981594, 0.99981756, 0.99982017, 0.99982329,\n",
       "        0.99982785, 0.99982841, 0.9998481 , 0.99986325, 0.9998675 ,\n",
       "        0.99987109, 0.99987644, 0.99988556, 0.99989129, 0.99989295,\n",
       "        0.99989342, 0.99989349, 0.99989788, 0.99990197, 0.99990255,\n",
       "        0.99990632, 0.99991193, 0.99991732, 0.99991887, 0.99992079,\n",
       "        0.99992097, 0.99992248, 0.99993048, 0.99993223, 0.99994658,\n",
       "        0.99994678, 0.99995296, 0.99995348, 0.99996042, 0.99996065,\n",
       "        0.99996961, 0.9999709 , 0.99997142, 0.99997492, 0.99997665,\n",
       "        0.9999783 , 0.99998037, 0.99998073, 0.99998436, 0.99998438,\n",
       "        0.99998635, 0.99998918, 0.99998958, 0.99998966, 0.9999901 ,\n",
       "        0.9999923 , 0.9999932 , 0.99999404, 0.9999942 , 0.99999438,\n",
       "        0.9999946 , 0.9999953 , 0.99999609, 0.99999654, 0.99999784,\n",
       "        0.99999798, 0.99999827, 0.99999838, 0.99999851, 0.99999949,\n",
       "        0.99999955, 0.99999995]))"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# EJERCICIO: calcular los puntos de la curva Precision-Recall mediante el método proporcionado por Sklearn\n",
    "metrics.precision_recall_curve(y, y_proba_clase1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Prec-Recall curve')"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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: mostrar gráficamente la curva Precision-Recall\n",
    "from sklearn.utils.fixes import signature\n",
    "\n",
    "precision, recall, _ = metrics.precision_recall_curve(y, y_proba_clase1)\n",
    "\n",
    "step_kwargs = ({'step': 'post'}\n",
    "               if 'step' in signature(plt.fill_between).parameters\n",
    "               else {})\n",
    "plt.step(recall, precision, color='b', alpha=0.2,\n",
    "         where='post')\n",
    "plt.fill_between(recall, precision, alpha=0.2, color='b', **step_kwargs)\n",
    "\n",
    "plt.xlabel('Recall')\n",
    "plt.ylabel('Precision')\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.title('Prec-Recall curve')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9942337916781663\n"
     ]
    }
   ],
   "source": [
    "# EJERCICIO: calcular la métrica average_precision_score mediante el método proporcionado por Sklearn\n",
    "print(metrics.average_precision_score(y, y_proba_clase1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. Otros clasificadores <a name=\"otros_clasif\"> </a>\n",
    "\n",
    "### 6.1. Árboles de Decisión <a name=\"arboles\"> </a>\n",
    "\n",
    "Un árbol de decisión es una estructura que nos permite responder a preguntas complejas, mediante una secuencia ordenada de preguntas simples.\n",
    "\n",
    "La clasificación automática basada en árboles de decisión consiste en la creación de un árbol de este tipo a partir de un conjunto de datos. Es uno de los métodos más versátiles y populares. Tiene, además, la ventaja de que el modelo que se construye es interpretable por las personas.\n",
    "\n",
    "#### 6.1.1. Análisis del dataset con Árboles de Decisión"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: Repetir el análisis anterior haciendo uso del clasificador DecisionTreeClassifier\n",
    "clf = DecisionTreeClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 91.3884007029877%\n",
      "Confusion matrix: [[194  18]\n",
      " [ 31 326]]\n"
     ]
    }
   ],
   "source": [
    "get_metrics(clf, X, y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TPR: 0.9215686274509803\n",
      "FPR: 0.09905660377358491\n",
      "ROC Curve: (array([0.        , 0.08490566, 1.        ]), array([0.        , 0.91596639, 1.        ]), array([2., 1., 0.]))\n",
      "AUC ROC: 0.9155303630886316\n"
     ]
    },
    {
     "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: Análisis ROC \n",
    "get_metrics2(clf, X, y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "#### 6.1.2 Interpretación de resultados con Árboles de Decisión\n",
    "\n",
    "Algunos modelos asociados a algoritmos de aprendizaje (como ocurre con los árboles de decisión) son interpretables. Esto significa que del modelo se pueden extraer reglas, o patrones, comprensibles para las personas. En este ejercicio usaremos una característica del estimador <code>DecisionTreeClassifier</code> que nos permite visualizar el árbol construido tras ser entrenado. \n",
    "\n",
    "Para ello nos apoyaremos en el paquete <code>graphviz</code> que permite visualizar grafos. La instrucción para instalar <code>graphviz</code> desde la consola de <code>anaconda</code> es:\n",
    "- <code>conda install python-graphviz</code>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "# EJERCICIO: Visualizar los dos primeros niveles del árbol resultante de entrenar el estimador DecisionTreeClassifier con los datos X e y.\n",
    "# nota: hacer uso de Source y tree.export_graphviz\n",
    "from IPython.display import Image  \n",
    "\n",
    "clf = DecisionTreeClassifier()\n",
    "clf.fit(X, y)\n",
    "\n",
    "dot_data = tree.export_graphviz(clf,\n",
    "                                out_file=None,\n",
    "                                filled=True,\n",
    "                                rounded=True)\n",
    "graph = pydotplus.graph_from_dot_data(dot_data)\n",
    "\n",
    "#Image(graph.create_png())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6.2. Random Forest <a name=\"randomF\"> </a>\n",
    "\n",
    "Random forest es un método que combina una cantidad grande de árboles de decisión independientes probados sobre conjuntos de datos aleatorios con igual distribución."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 96.30931458699473%\n",
      "Confusion matrix: [[199  13]\n",
      " [  8 349]]\n",
      "TPR: 0.9775910364145658\n",
      "FPR: 0.06132075471698113\n",
      "ROC Curve: (array([0.        , 0.00471698, 0.00471698, 0.00943396, 0.01415094,\n",
      "       0.01415094, 0.01415094, 0.01415094, 0.01415094, 0.01415094,\n",
      "       0.01415094, 0.01415094, 0.01415094, 0.01415094, 0.01415094,\n",
      "       0.01886792, 0.01886792, 0.01886792, 0.01886792, 0.01886792,\n",
      "       0.01886792, 0.01886792, 0.02830189, 0.02830189, 0.03301887,\n",
      "       0.03301887, 0.04245283, 0.04245283, 0.05188679, 0.05188679,\n",
      "       0.05660377, 0.07075472, 0.08018868, 0.08018868, 0.08962264,\n",
      "       0.08962264, 0.0990566 , 0.10849057, 0.11320755, 0.13207547,\n",
      "       0.13679245, 0.14150943, 0.14150943, 0.14622642, 0.16037736,\n",
      "       0.16509434, 0.18396226, 0.19811321, 0.20754717, 0.21698113,\n",
      "       0.23584906, 0.25      , 0.25943396, 0.27358491, 0.29245283,\n",
      "       0.30660377, 0.3254717 , 0.3490566 , 0.36320755, 0.4009434 ,\n",
      "       0.43396226, 0.48584906, 0.58490566, 1.        ]), array([0.        , 0.42577031, 0.54901961, 0.64705882, 0.69467787,\n",
      "       0.74509804, 0.76470588, 0.77591036, 0.80112045, 0.80672269,\n",
      "       0.82072829, 0.83753501, 0.8487395 , 0.85154062, 0.85994398,\n",
      "       0.87114846, 0.88235294, 0.8907563 , 0.89635854, 0.89915966,\n",
      "       0.91036415, 0.91316527, 0.91876751, 0.94677871, 0.94957983,\n",
      "       0.95238095, 0.95518207, 0.96078431, 0.96358543, 0.9719888 ,\n",
      "       0.9719888 , 0.97478992, 0.97759104, 0.98319328, 0.98319328,\n",
      "       0.98879552, 0.99159664, 0.99159664, 0.99159664, 0.99159664,\n",
      "       0.99159664, 0.99439776, 0.99719888, 0.99719888, 0.99719888,\n",
      "       0.99719888, 0.99719888, 0.99719888, 1.        , 1.        ,\n",
      "       1.        , 1.        , 1.        , 1.        , 1.        ,\n",
      "       1.        , 1.        , 1.        , 1.        , 1.        ,\n",
      "       1.        , 1.        , 1.        , 1.        ]), array([2.  , 1.  , 0.99, 0.98, 0.97, 0.95, 0.94, 0.93, 0.92, 0.91, 0.9 ,\n",
      "       0.88, 0.87, 0.86, 0.85, 0.84, 0.81, 0.8 , 0.79, 0.78, 0.75, 0.74,\n",
      "       0.73, 0.67, 0.66, 0.65, 0.64, 0.62, 0.58, 0.53, 0.52, 0.51, 0.48,\n",
      "       0.46, 0.45, 0.44, 0.4 , 0.39, 0.38, 0.35, 0.32, 0.31, 0.27, 0.26,\n",
      "       0.25, 0.24, 0.21, 0.18, 0.17, 0.15, 0.13, 0.12, 0.11, 0.1 , 0.09,\n",
      "       0.08, 0.07, 0.06, 0.05, 0.04, 0.03, 0.02, 0.01, 0.  ]))\n",
      "AUC ROC: 0.988868188784948\n"
     ]
    },
    {
     "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: Repetir el análisis anterior haciendo uso del clasificador RandomForestClassifier\n",
    "clf = RandomForestClassifier(n_estimators=100)\n",
    "\n",
    "get_metrics(clf, X, y)\n",
    "\n",
    "get_metrics2(clf, X, y)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}