p1_bcw.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import os\n",
    "from sklearn import datasets\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import itertools as it\n",
    "from sklearn.neighbors import LocalOutlierFactor\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV, KFold\n",
    "from sklearn.discriminant_analysis import (LinearDiscriminantAnalysis,\n",
    "                                           QuadraticDiscriminantAnalysis)\n",
    "from sklearn.metrics import confusion_matrix, classification_report\n",
    "from sklearn.utils import shuffle\n",
    "from scipy import stats\n",
    "from statsmodels.stats.multicomp import (pairwise_tukeyhsd,\n",
    "                                         MultiComparison)\n",
    "from statsmodels.sandbox.stats.multicomp import TukeyHSDResults\n",
    "from sklearn.metrics import roc_curve\n",
    "from sklearn import preprocessing\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.model_selection import StratifiedKFold, cross_val_score\n",
    "from sklearn.utils.class_weight import compute_class_weight\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\", category=DeprecationWarning)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cargamos dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Cargamos dataset Breast Cancer Wisconsin (Original)\n",
    "ds = pd.read_csv('./breast-cancer-wisconsin.data', usecols=[1,2,3,4,5,6,7,8,9,10], \n",
    "                 names=[0,1,2,3,4,5,6,7,8,9], header=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(699, 10)"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ds.shape\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "output = 9\n",
    "# Transformamos la columna de clases a 0-1 (actualmente son 2,4)\n",
    "ds[output] = np.uint8(ds[output]/2-1)\n",
    "#y = ds[output].values\n",
    "#X = ds.drop(output,axis=1).values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocesado de datos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Hacemos undersampling para intentar mejorar los resultados**  \n",
    "Ya que después de unas pruebas preliminares con el dataset desbalanceado obteníamos mucha diferencia entre Sensibilidad y Especificidad."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "ds_u = ds.copy()\n",
    "# Número de casos positivos\n",
    "nposit = len(ds_u[ds_u[output] == 1])\n",
    "# Índices de casos negativos\n",
    "inegat = ds_u[ds_u[output] == 0].index\n",
    "# Cogemos aleatoriamente un número de índices negativos\n",
    "# igual al número de casos positivos\n",
    "random_indices = np.random.choice(inegat, nposit, replace=False)\n",
    "# Obtenemos todos los índices positivos\n",
    "iposit = ds_u[ds_u[output] == 1].index\n",
    "# Unimos las dos listas de índices\n",
    "new_indices = np.concatenate([iposit, random_indices])\n",
    "ds_u = ds_u.loc[new_indices]\n",
    "# y las desordenamos\n",
    "ds_u = shuffle(ds_u).reset_index(drop=True)\n",
    "\n",
    "X_u = ds_u.drop(output, axis=1)\n",
    "y_u = ds_u[output]\n",
    "\n",
    "X = X_u.values\n",
    "y = y_u.values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Comprobamos que no haya \"missing values\"**  \n",
    "Según https://archive.ics.uci.edu/ml/datasets, los Missing Values están etiquetados como '?'.  \n",
    "Eliminamos las filas con dicho valor."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "rows = np.argwhere(X == '?')[:,0]\n",
    "X = np.delete(X,rows,0)\n",
    "y = np.delete(y,rows)\n",
    "X = X.astype(np.uint8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "attr = []\n",
    "# Iteramos los atributos del dataset\n",
    "for i in ds.columns:\n",
    "    # Comprobamos los valores de cada atributo\n",
    "    attr.append(np.all(ds[i].notna()))\n",
    "# Ningún atributo tiene missing values:\n",
    "np.all(np.array(attr))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Comprobamos que no haya variables con \"near-zero variance\"**  \n",
    "Vemos que no hay ninguna."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Int64Index([], dtype='int64')"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ds.columns.drop(output)[ds.std() < 0.001]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Comprobamos que no haya variables iguales**  \n",
    "Vemos que no hay ninguna."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "resultados = np.zeros((len(ds.columns), len(ds.columns)))\n",
    "for i in range(len(ds.columns)):\n",
    "    for j in range(len(ds.columns)):\n",
    "        resultados[i, j] = np.mean(ds.iloc[:, i] == ds.iloc[:, j])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "iguales = []\n",
    "for i, j in it.combinations(range(len(ds.columns)), 2):\n",
    "    if round(resultados[i, j], 4) == 1:\n",
    "        iguales.append(ds.columns[i])\n",
    "        print(ds.columns[i], ds.columns[j])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Comprobamos que las clases no estén desbalanceadas**  \n",
    "Ya no están desbalanceadas ya que hemos hecho undersampling."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Proporción clase 0:  0.4936440677966102\n",
      "Proporción clase 1:  0.5063559322033898\n"
     ]
    }
   ],
   "source": [
    "print('Proporción clase 0: ',np.mean(y==0))\n",
    "print('Proporción clase 1: ',np.mean(y==1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Desordenamos los datos**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "X,y = shuffle(X,y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Detección de Outliers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = LocalOutlierFactor(n_neighbors=20, contamination=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = clf.fit_predict(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_scores = clf.negative_outlier_factor_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Este procedimiento no tiene mucho sentido con valores discretos**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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": [
    "plt.title(\"Local Outlier Factor (LOF)\")\n",
    "plt.scatter(X[:, 0], X[:, 1], color='k', s=3., label='Data points')\n",
    "# plot circles with radius proportional to the outlier scores\n",
    "radius = (X_scores.max() - X_scores) / (X_scores.max() - X_scores.min())\n",
    "plt.scatter(X[:, 0], X[:, 1], s=1000 * radius, edgecolors='r',\n",
    "            facecolors='none', label='Outlier scores')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, ax = plt.subplots(2,1,figsize=(9,5))\n",
    "for i in set(y):\n",
    "    temp = X[y == i]\n",
    "    y_pred = clf.fit_predict(temp)\n",
    "    temp_scores = clf.negative_outlier_factor_\n",
    "    ax[i].plot(range(len(temp_scores)),sorted(-temp_scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, ax = plt.subplots(1,2,figsize=(15,5))\n",
    "for i in set(y):\n",
    "    temp = X[y == i]\n",
    "    data = []\n",
    "    for j in range(X.shape[1]):\n",
    "        data.append(temp[:,j])\n",
    "    bp = ax[i].boxplot(data)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create a figure instance\n",
    "fig = plt.figure(figsize=(8, 5))\n",
    "# Create an axes instance\n",
    "ax = fig.add_subplot(111)\n",
    "data = []\n",
    "for j in range(X.shape[1]):\n",
    "    data.append(X[:,j])\n",
    "bp = ax.boxplot(data)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Entrenamiento de Modelos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**kFold - Linear Discriminant Analysis**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "kf = KFold(n_splits=10)\n",
    "resultados_train = []\n",
    "resultados_test = []\n",
    "for tipo in ('lsqr', 'eigen', 'svd'):\n",
    "    for _ in range(3):\n",
    "        for train_index, test_index in kf.split(X):\n",
    "            X_train, X_test = X[train_index], X[test_index]\n",
    "            y_train, y_test = y[train_index], y[test_index]\n",
    "            alg = LinearDiscriminantAnalysis(solver=tipo, shrinkage=None)\n",
    "            alg.fit(X_train, y_train)\n",
    "            resultados_train.append(alg.score(X_train, y_train))\n",
    "            resultados_test.append(alg.score(X_test, y_test))\n",
    "resultados_train = np.reshape(resultados_train,(30,3),order='F')\n",
    "resultados_test = np.reshape(resultados_test,(30,3),order='F')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Test de normalidad**  \n",
    "Rechazamos normalidad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0010928306728601456\n",
      "0.0010928306728601456\n",
      "0.0010928306728601456\n"
     ]
    }
   ],
   "source": [
    "print(stats.shapiro(resultados_test[:,0])[1])\n",
    "print(stats.shapiro(resultados_test[:,1])[1])\n",
    "print(stats.shapiro(resultados_test[:,2])[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Test de Kruskal-Wallis**  \n",
    "No se puede rechazar que las precisiones sean similares"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "KruskalResult(statistic=0.0, pvalue=1.0)\n"
     ]
    }
   ],
   "source": [
    "print(stats.kruskal(resultados_test[:,0],\n",
    "                    resultados_test[:,1],\n",
    "                    resultados_test[:,2]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**kFold - Quadratic Discriminant Analysis**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# kf = KFold(n_splits=10)\n",
    "kf = KFold(n_splits=5)\n",
    "resultados_trainq = []\n",
    "resultados_testq = []\n",
    "# for _ in range(3):\n",
    "for _ in range(6):\n",
    "    for train_index, test_index in kf.split(X):\n",
    "        X_train, X_test = X[train_index], X[test_index]\n",
    "        y_train, y_test = y[train_index], y[test_index]\n",
    "        alg = QuadraticDiscriminantAnalysis(priors=np.array([1/3, 1/3]))\n",
    "        alg.fit(X_train, y_train)\n",
    "        resultados_trainq.append(alg.score(X_train, y_train))\n",
    "        resultados_testq.append(alg.score(X_test, y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Test de normalidad**  \n",
    "Rechazamos normalidad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.001521045807749033\n"
     ]
    }
   ],
   "source": [
    "print(stats.shapiro(resultados_testq)[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Test de Kruskal-Wallis**  \n",
    "No se puede rechazar que las precisiones sean similares"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "KruskalResult(statistic=0.2536161815490905, pvalue=0.9685021196313555)\n"
     ]
    }
   ],
   "source": [
    "print(stats.kruskal(resultados_test[:,0],\n",
    "                    resultados_test[:,1],\n",
    "                    resultados_test[:,2],\n",
    "                    resultados_testq))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Tukey**  \n",
    "Ya que no podemos rechazar que las precisiones sean similares,  \n",
    "este test no es necesario (sin embargo, sin hacer el undersampling sí haría falta)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "resultados_test = resultados_test.flatten(\"F\")\n",
    "resultados_testq = np.array(resultados_testq)\n",
    "resultados_test = np.concatenate((resultados_test,resultados_testq))\n",
    "labels = [i for i in \"0123\" for _ in range(int(resultados_test.shape[0]/4))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "resultados = pairwise_tukeyhsd(resultados_test,labels)\n",
    "resultados.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Figure(720x432)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(resultados.plot_simultaneous())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Hold Out**  \n",
    "//TODO: mover este cálculo a cada iteración del k-fold.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "def performance_indexes(CM, PositiveClass):\n",
    "    classNum = len(CM)\n",
    "    TP = CM[PositiveClass, PositiveClass]\n",
    "    TN = 0\n",
    "    FP = 0\n",
    "    FN = 0\n",
    "    for real in range(classNum):\n",
    "        for predicted in range(classNum):\n",
    "            if (real != PositiveClass and predicted != PositiveClass):\n",
    "                TN += CM[real, predicted]\n",
    "            if (real != PositiveClass and predicted == PositiveClass):\n",
    "                FP += CM[real, predicted]\n",
    "            if (real == PositiveClass and predicted != PositiveClass):\n",
    "                FN += CM[real, predicted]\n",
    "    Sens = TP/(TP+FN)\n",
    "    Spec = TN/(TN+FP)\n",
    "    PPV = TP/(TP+FP)\n",
    "    NPV = TN/(TN+FN)\n",
    "    ACC = (TP+TN)/(TP+TN+FP+FN)\n",
    "    return Sens,Spec,PPV,NPV,ACC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sens: 0.98, Spec: 0.93, PPV: 0.94, NPV: 0.98, ACC: 0.96\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.9576271186440678"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.25)\n",
    "alg = LinearDiscriminantAnalysis(solver='lsqr', shrinkage=None)\n",
    "alg.fit(X_train, y_train)\n",
    "y_pred = alg.predict(X_test)\n",
    "CM = confusion_matrix(y_test, y_pred)\n",
    "Sens,Spec,PPV,NPV,ACC = performance_indexes(CM,0)\n",
    "print(('Sens: {:.2f}, Spec: {:.2f}, PPV: {:.2f}, ' +\n",
    "      'NPV: {:.2f}, ACC: {:.2f}').format(Sens,Spec,PPV,NPV,ACC))\n",
    "alg.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sens: 0.95, Spec: 0.98, PPV: 0.98, NPV: 0.95, ACC: 0.97\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.9661016949152542"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.25)\n",
    "alg = QuadraticDiscriminantAnalysis(priors=np.array([1/3, 1/3]))\n",
    "alg.fit(X_train, y_train)\n",
    "y_pred = alg.predict(X_test)\n",
    "CM = confusion_matrix(y_test, y_pred)\n",
    "Sens,Spec,PPV,NPV,ACC = performance_indexes(CM,0)\n",
    "print(('Sens: {:.2f}, Spec: {:.2f}, PPV: {:.2f}, ' +\n",
    "      'NPV: {:.2f}, ACC: {:.2f}').format(Sens,Spec,PPV,NPV,ACC))\n",
    "alg.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**----------------------------------------------------------------------------------------------------------------------------**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Persistencia"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# with open('RForest_Zika_oversampled.bin','wb') as file:\n",
    "#     pickle.dump(alg_o, file)\n",
    "# with open('RForest_Zika_undersampled.bin','wb') as file:\n",
    "#     pickle.dump(alg_u, file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## FIN"
   ]
  }
 ],
 "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
}