ClasifEEG.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Adquisición de los datos\n",
    "Descarga de los datos. Hago el GET y lo copio en un archivo 'EEGEyeState.arff'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os.path\n",
    "import urllib.request\n",
    "import shutil\n",
    "\n",
    "url = 'https://archive.ics.uci.edu/ml/machine-learning-databases/00264/EEG%20Eye%20State.arff'\n",
    "file_name = 'EEGEyeState.arff'\n",
    "\n",
    "# Si el archivo no existe o tiene tamaño 0, lo bajo\n",
    "if not os.path.exists(file_name) or os.stat(file_name).st_size <= 0:\n",
    "    with urllib.request.urlopen(url) as response, open(file_name, 'wb') as out_file:\n",
    "        shutil.copyfileobj(response, out_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Analisis inicial y detección de outliers"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Cargo los datos del archivo arff y los convierto a dataframe.\n",
    "Hago describe para dar un vistazo a los datos. \n",
    "Los máximos de algunas variables (AF3, FC5, P7, O1, P8, F8, AF4) están sumamente lejos de las medias. Es probable que hayan outliers en el dataset.\n",
    "La media de eyeDetection está cercana a 0.5 así que es un dataset balanceado (i.e: no hay mucho más etiquetas de una categoría que de otra)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>AF3</th>\n",
       "      <th>F7</th>\n",
       "      <th>F3</th>\n",
       "      <th>FC5</th>\n",
       "      <th>T7</th>\n",
       "      <th>P7</th>\n",
       "      <th>O1</th>\n",
       "      <th>O2</th>\n",
       "      <th>P8</th>\n",
       "      <th>T8</th>\n",
       "      <th>FC6</th>\n",
       "      <th>F4</th>\n",
       "      <th>F8</th>\n",
       "      <th>AF4</th>\n",
       "      <th>eyeDetection</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "      <td>14980.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>4321.917777</td>\n",
       "      <td>4009.767694</td>\n",
       "      <td>4264.022433</td>\n",
       "      <td>4164.946326</td>\n",
       "      <td>4341.741075</td>\n",
       "      <td>4644.022379</td>\n",
       "      <td>4110.400160</td>\n",
       "      <td>4616.056904</td>\n",
       "      <td>4218.826610</td>\n",
       "      <td>4231.316200</td>\n",
       "      <td>4202.456900</td>\n",
       "      <td>4279.232774</td>\n",
       "      <td>4615.205336</td>\n",
       "      <td>4416.435832</td>\n",
       "      <td>0.448798</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>2492.072174</td>\n",
       "      <td>45.941672</td>\n",
       "      <td>44.428052</td>\n",
       "      <td>5216.404632</td>\n",
       "      <td>34.738821</td>\n",
       "      <td>2924.789537</td>\n",
       "      <td>4600.926543</td>\n",
       "      <td>29.292603</td>\n",
       "      <td>2136.408523</td>\n",
       "      <td>38.050903</td>\n",
       "      <td>37.785981</td>\n",
       "      <td>41.544312</td>\n",
       "      <td>1208.369958</td>\n",
       "      <td>5891.285043</td>\n",
       "      <td>0.497388</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1030.770000</td>\n",
       "      <td>2830.770000</td>\n",
       "      <td>1040.000000</td>\n",
       "      <td>2453.330000</td>\n",
       "      <td>2089.740000</td>\n",
       "      <td>2768.210000</td>\n",
       "      <td>2086.150000</td>\n",
       "      <td>4567.180000</td>\n",
       "      <td>1357.950000</td>\n",
       "      <td>1816.410000</td>\n",
       "      <td>3273.330000</td>\n",
       "      <td>2257.950000</td>\n",
       "      <td>86.666700</td>\n",
       "      <td>1366.150000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>4280.510000</td>\n",
       "      <td>3990.770000</td>\n",
       "      <td>4250.260000</td>\n",
       "      <td>4108.210000</td>\n",
       "      <td>4331.790000</td>\n",
       "      <td>4611.790000</td>\n",
       "      <td>4057.950000</td>\n",
       "      <td>4604.620000</td>\n",
       "      <td>4190.770000</td>\n",
       "      <td>4220.510000</td>\n",
       "      <td>4190.260000</td>\n",
       "      <td>4267.690000</td>\n",
       "      <td>4590.770000</td>\n",
       "      <td>4342.050000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>4294.360000</td>\n",
       "      <td>4005.640000</td>\n",
       "      <td>4262.560000</td>\n",
       "      <td>4120.510000</td>\n",
       "      <td>4338.970000</td>\n",
       "      <td>4617.950000</td>\n",
       "      <td>4070.260000</td>\n",
       "      <td>4613.330000</td>\n",
       "      <td>4199.490000</td>\n",
       "      <td>4229.230000</td>\n",
       "      <td>4200.510000</td>\n",
       "      <td>4276.920000</td>\n",
       "      <td>4603.080000</td>\n",
       "      <td>4354.870000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>4311.790000</td>\n",
       "      <td>4023.080000</td>\n",
       "      <td>4270.770000</td>\n",
       "      <td>4132.310000</td>\n",
       "      <td>4347.180000</td>\n",
       "      <td>4626.670000</td>\n",
       "      <td>4083.590000</td>\n",
       "      <td>4624.100000</td>\n",
       "      <td>4209.230000</td>\n",
       "      <td>4239.490000</td>\n",
       "      <td>4211.280000</td>\n",
       "      <td>4287.180000</td>\n",
       "      <td>4617.440000</td>\n",
       "      <td>4372.820000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>309231.000000</td>\n",
       "      <td>7804.620000</td>\n",
       "      <td>6880.510000</td>\n",
       "      <td>642564.000000</td>\n",
       "      <td>6474.360000</td>\n",
       "      <td>362564.000000</td>\n",
       "      <td>567179.000000</td>\n",
       "      <td>7264.100000</td>\n",
       "      <td>265641.000000</td>\n",
       "      <td>6674.360000</td>\n",
       "      <td>6823.080000</td>\n",
       "      <td>7002.560000</td>\n",
       "      <td>152308.000000</td>\n",
       "      <td>715897.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 AF3            F7            F3            FC5            T7  \\\n",
       "count   14980.000000  14980.000000  14980.000000   14980.000000  14980.000000   \n",
       "mean     4321.917777   4009.767694   4264.022433    4164.946326   4341.741075   \n",
       "std      2492.072174     45.941672     44.428052    5216.404632     34.738821   \n",
       "min      1030.770000   2830.770000   1040.000000    2453.330000   2089.740000   \n",
       "25%      4280.510000   3990.770000   4250.260000    4108.210000   4331.790000   \n",
       "50%      4294.360000   4005.640000   4262.560000    4120.510000   4338.970000   \n",
       "75%      4311.790000   4023.080000   4270.770000    4132.310000   4347.180000   \n",
       "max    309231.000000   7804.620000   6880.510000  642564.000000   6474.360000   \n",
       "\n",
       "                  P7             O1            O2             P8  \\\n",
       "count   14980.000000   14980.000000  14980.000000   14980.000000   \n",
       "mean     4644.022379    4110.400160   4616.056904    4218.826610   \n",
       "std      2924.789537    4600.926543     29.292603    2136.408523   \n",
       "min      2768.210000    2086.150000   4567.180000    1357.950000   \n",
       "25%      4611.790000    4057.950000   4604.620000    4190.770000   \n",
       "50%      4617.950000    4070.260000   4613.330000    4199.490000   \n",
       "75%      4626.670000    4083.590000   4624.100000    4209.230000   \n",
       "max    362564.000000  567179.000000   7264.100000  265641.000000   \n",
       "\n",
       "                 T8           FC6            F4             F8            AF4  \\\n",
       "count  14980.000000  14980.000000  14980.000000   14980.000000   14980.000000   \n",
       "mean    4231.316200   4202.456900   4279.232774    4615.205336    4416.435832   \n",
       "std       38.050903     37.785981     41.544312    1208.369958    5891.285043   \n",
       "min     1816.410000   3273.330000   2257.950000      86.666700    1366.150000   \n",
       "25%     4220.510000   4190.260000   4267.690000    4590.770000    4342.050000   \n",
       "50%     4229.230000   4200.510000   4276.920000    4603.080000    4354.870000   \n",
       "75%     4239.490000   4211.280000   4287.180000    4617.440000    4372.820000   \n",
       "max     6674.360000   6823.080000   7002.560000  152308.000000  715897.000000   \n",
       "\n",
       "       eyeDetection  \n",
       "count  14980.000000  \n",
       "mean       0.448798  \n",
       "std        0.497388  \n",
       "min        0.000000  \n",
       "25%        0.000000  \n",
       "50%        0.000000  \n",
       "75%        1.000000  \n",
       "max        1.000000  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from scipy.io import arff\n",
    "import pandas as pd\n",
    "data, meta = arff.loadarff(file_name)\n",
    "data = pd.DataFrame(data)\n",
    "# Convierto la columna final a entero\n",
    "data[\"eyeDetection\"] = data[\"eyeDetection\"].astype('int')\n",
    "data.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Aplico 3 técnicas de detección de outliers no paramétricas (a priori no sabemos nada de las variables que componen los datos) aptas para datos de alta dimensión.\n",
    "One Class SVM, Isolation Forest y Local Outlier Factor. Encontradas en: http://scikit-learn.org/stable/modules/outlier_detection.html\n",
    "Clasifico un dato como outlier si 2 de los 3 métodos lo detectan como tal."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn import svm\n",
    "from sklearn.ensemble import IsolationForest\n",
    "from sklearn.neighbors import LocalOutlierFactor\n",
    "\n",
    "# Semilla aleatoria para resultados reproducibles\n",
    "rng = np.random.RandomState(42)\n",
    "factorContaminacion = 0.5/100 # Arbitrario pero conservador, 0.5% de los datos se consideran outliers, se podría ajustar mejor\n",
    "\n",
    "cOneClassSVM = svm.OneClassSVM(nu=1 - factorContaminacion, kernel=\"rbf\", gamma=0.1, random_state=rng)\n",
    "cIsolationForest = IsolationForest(contamination=factorContaminacion, random_state=rng, n_jobs=-1)\n",
    "cLocalOutlierFactor = LocalOutlierFactor(n_neighbors=35, contamination=factorContaminacion, n_jobs=-1)\n",
    "\n",
    "# Ajusta los modelos y calcula las etiquetas de cada dato para cada método. Demora unos minutos\n",
    "cOneClassSVM.fit(X=data)\n",
    "cIsolationForest.fit(X=data)\n",
    "esOutlier = np.column_stack((cOneClassSVM.predict(X=data), cIsolationForest.predict(X=data), cLocalOutlierFactor.fit_predict(X=data))) == -1\n",
    "esOutlier = esOutlier.sum(axis=1) >= 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "19 valores fueron detectados como outliers por la combinación de métodos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "19"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "esOutlier.sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Quitando los outliers los máximos se ven más razonables. Además, las desviaciones de las variables afectadas disminuyen notoriamente."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>AF3</th>\n",
       "      <th>F7</th>\n",
       "      <th>F3</th>\n",
       "      <th>FC5</th>\n",
       "      <th>T7</th>\n",
       "      <th>P7</th>\n",
       "      <th>O1</th>\n",
       "      <th>O2</th>\n",
       "      <th>P8</th>\n",
       "      <th>T8</th>\n",
       "      <th>FC6</th>\n",
       "      <th>F4</th>\n",
       "      <th>F8</th>\n",
       "      <th>AF4</th>\n",
       "      <th>eyeDetection</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "      <td>14961.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>4301.284253</td>\n",
       "      <td>4009.425433</td>\n",
       "      <td>4263.989699</td>\n",
       "      <td>4122.285475</td>\n",
       "      <td>4341.471710</td>\n",
       "      <td>4619.957046</td>\n",
       "      <td>4072.741916</td>\n",
       "      <td>4615.703193</td>\n",
       "      <td>4201.255304</td>\n",
       "      <td>4231.107323</td>\n",
       "      <td>4202.053115</td>\n",
       "      <td>4278.997516</td>\n",
       "      <td>4605.915226</td>\n",
       "      <td>4361.214690</td>\n",
       "      <td>0.449302</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>37.657989</td>\n",
       "      <td>29.974471</td>\n",
       "      <td>21.091153</td>\n",
       "      <td>20.701946</td>\n",
       "      <td>16.724244</td>\n",
       "      <td>18.166280</td>\n",
       "      <td>20.737372</td>\n",
       "      <td>18.110231</td>\n",
       "      <td>17.599426</td>\n",
       "      <td>19.500633</td>\n",
       "      <td>24.246670</td>\n",
       "      <td>19.854582</td>\n",
       "      <td>33.271561</td>\n",
       "      <td>38.374001</td>\n",
       "      <td>0.497440</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>4197.950000</td>\n",
       "      <td>3905.640000</td>\n",
       "      <td>4197.440000</td>\n",
       "      <td>4058.460000</td>\n",
       "      <td>4304.620000</td>\n",
       "      <td>4566.150000</td>\n",
       "      <td>4026.150000</td>\n",
       "      <td>4567.180000</td>\n",
       "      <td>4147.690000</td>\n",
       "      <td>4152.820000</td>\n",
       "      <td>4100.000000</td>\n",
       "      <td>4201.030000</td>\n",
       "      <td>4443.080000</td>\n",
       "      <td>4205.640000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>4280.510000</td>\n",
       "      <td>3990.770000</td>\n",
       "      <td>4250.260000</td>\n",
       "      <td>4108.210000</td>\n",
       "      <td>4331.790000</td>\n",
       "      <td>4611.790000</td>\n",
       "      <td>4057.950000</td>\n",
       "      <td>4604.620000</td>\n",
       "      <td>4190.770000</td>\n",
       "      <td>4220.510000</td>\n",
       "      <td>4190.260000</td>\n",
       "      <td>4267.690000</td>\n",
       "      <td>4590.770000</td>\n",
       "      <td>4342.050000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>4294.360000</td>\n",
       "      <td>4005.640000</td>\n",
       "      <td>4262.560000</td>\n",
       "      <td>4120.510000</td>\n",
       "      <td>4338.970000</td>\n",
       "      <td>4617.950000</td>\n",
       "      <td>4070.260000</td>\n",
       "      <td>4613.330000</td>\n",
       "      <td>4199.490000</td>\n",
       "      <td>4229.230000</td>\n",
       "      <td>4200.510000</td>\n",
       "      <td>4276.920000</td>\n",
       "      <td>4603.080000</td>\n",
       "      <td>4354.870000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>4311.790000</td>\n",
       "      <td>4023.080000</td>\n",
       "      <td>4270.770000</td>\n",
       "      <td>4132.310000</td>\n",
       "      <td>4347.180000</td>\n",
       "      <td>4626.670000</td>\n",
       "      <td>4083.590000</td>\n",
       "      <td>4624.100000</td>\n",
       "      <td>4209.230000</td>\n",
       "      <td>4239.490000</td>\n",
       "      <td>4211.280000</td>\n",
       "      <td>4287.180000</td>\n",
       "      <td>4617.440000</td>\n",
       "      <td>4372.820000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>4504.100000</td>\n",
       "      <td>4156.920000</td>\n",
       "      <td>4386.150000</td>\n",
       "      <td>4250.260000</td>\n",
       "      <td>4463.590000</td>\n",
       "      <td>4756.920000</td>\n",
       "      <td>4178.460000</td>\n",
       "      <td>4731.790000</td>\n",
       "      <td>4320.000000</td>\n",
       "      <td>4362.560000</td>\n",
       "      <td>4332.310000</td>\n",
       "      <td>4394.870000</td>\n",
       "      <td>4833.850000</td>\n",
       "      <td>4573.330000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                AF3            F7            F3           FC5            T7  \\\n",
       "count  14961.000000  14961.000000  14961.000000  14961.000000  14961.000000   \n",
       "mean    4301.284253   4009.425433   4263.989699   4122.285475   4341.471710   \n",
       "std       37.657989     29.974471     21.091153     20.701946     16.724244   \n",
       "min     4197.950000   3905.640000   4197.440000   4058.460000   4304.620000   \n",
       "25%     4280.510000   3990.770000   4250.260000   4108.210000   4331.790000   \n",
       "50%     4294.360000   4005.640000   4262.560000   4120.510000   4338.970000   \n",
       "75%     4311.790000   4023.080000   4270.770000   4132.310000   4347.180000   \n",
       "max     4504.100000   4156.920000   4386.150000   4250.260000   4463.590000   \n",
       "\n",
       "                 P7            O1            O2            P8            T8  \\\n",
       "count  14961.000000  14961.000000  14961.000000  14961.000000  14961.000000   \n",
       "mean    4619.957046   4072.741916   4615.703193   4201.255304   4231.107323   \n",
       "std       18.166280     20.737372     18.110231     17.599426     19.500633   \n",
       "min     4566.150000   4026.150000   4567.180000   4147.690000   4152.820000   \n",
       "25%     4611.790000   4057.950000   4604.620000   4190.770000   4220.510000   \n",
       "50%     4617.950000   4070.260000   4613.330000   4199.490000   4229.230000   \n",
       "75%     4626.670000   4083.590000   4624.100000   4209.230000   4239.490000   \n",
       "max     4756.920000   4178.460000   4731.790000   4320.000000   4362.560000   \n",
       "\n",
       "                FC6            F4            F8           AF4  eyeDetection  \n",
       "count  14961.000000  14961.000000  14961.000000  14961.000000  14961.000000  \n",
       "mean    4202.053115   4278.997516   4605.915226   4361.214690      0.449302  \n",
       "std       24.246670     19.854582     33.271561     38.374001      0.497440  \n",
       "min     4100.000000   4201.030000   4443.080000   4205.640000      0.000000  \n",
       "25%     4190.260000   4267.690000   4590.770000   4342.050000      0.000000  \n",
       "50%     4200.510000   4276.920000   4603.080000   4354.870000      0.000000  \n",
       "75%     4211.280000   4287.180000   4617.440000   4372.820000      1.000000  \n",
       "max     4332.310000   4394.870000   4833.850000   4573.330000      1.000000  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "validData = data[~esOutlier]\n",
    "validData.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Mirando los histogramas las variables parecen razonablemente gaussianas. Vamos a usar escalado con media y desviación estandar"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[<matplotlib.axes._subplots.AxesSubplot object at 0x000002217A91EA20>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x000002217A77ECC0>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179E3A780>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179E5ADD8>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179E8D4A8>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179E8D4E0>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179EE6208>]\n",
      " [<matplotlib.axes._subplots.AxesSubplot object at 0x0000022179F0E898>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179F38F28>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179F695F8>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179F90C88>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179FC1358>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x0000022179FE99E8>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x000002217A01D0B8>]\n",
      " [<matplotlib.axes._subplots.AxesSubplot object at 0x000002217A044748>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x000002217A06CDD8>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x000002217A09D4A8>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x000002217A0C7B38>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x000002217A0F8208>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x000002217A121898>\n",
      "  <matplotlib.axes._subplots.AxesSubplot object at 0x000002217A149F28>]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2160x1080 with 21 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "print(validData.hist(layout=(3, 7), figsize=(30, 15)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conjuntos de entrenamiento y validación"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Divido los datos en un conjunto de entrenamiento (66.6% de la población) y validación (33.3%)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Labels are the values we want to predict\n",
    "labels = np.array(validData['eyeDetection'])\n",
    "\n",
    "# Remove the labels from the features\n",
    "# axis 1 refers to the columns\n",
    "features= validData.drop('eyeDetection', axis = 1)\n",
    "\n",
    "# Saving feature names for later use\n",
    "feature_list = list(features.columns)\n",
    "\n",
    "# Convert to numpy array\n",
    "features = np.array(features)\n",
    "\n",
    "# Split the data into training and testing sets\n",
    "train_features, test_features, train_labels, test_labels = train_test_split(features, labels, test_size = 1/3, random_state = rng)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Clasificador Binario con el Algoritmo K*\n",
    "En el paper en: http://suendermann.com/su/pdf/aihls2013.pdf el algoritmo que encuentra los mejores resultados es este K*, por lo que fue mi primer intento de solucionar el problema.\n",
    "Me cree un usuario en WEKA, conecté a la API con mi key, logré hacer que un llamado de ejemplo anduviera pero cuando lo voy a correr para el problema real me falla con un error de la clase genérica de la API (AlgorithmException) indicando que podría haber un timeout o un out-of-memory.\n",
    "Intenté aumentar el timeout a 50 minutos (el máximo que permite) pero no anduvo y la verdad no encontré mucho para donde seguir con el problema. No parecen ser muchos datos como para quedarse sin memoria. En este momento la página de la documentación del algoritmo en: http://weka.sourceforge.net/doc.stable/weka/classifiers/lazy/KStar.html me está dando un mensaje de error de que los servidores de sourceforge están experimentando un tráfico anormal, quizás pueda venir por ahí el problema."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "ename": "AlgorithmException",
     "evalue": "'Failed to run algorithm - may have timed-out or hit out-of-memory error'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mAlgorithmException\u001b[0m                        Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-64-47f676a61e62>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     13\u001b[0m \u001b[0malgo\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0malgo\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mset_options\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtimeout\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m3000\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     14\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 15\u001b[1;33m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0malgo\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpipe\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mresult\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32mC:\\ProgramData\\Anaconda3\\lib\\site-packages\\Algorithmia\\algorithm.py\u001b[0m in \u001b[0;36mpipe\u001b[1;34m(self, input1)\u001b[0m\n\u001b[0;32m     38\u001b[0m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_postVoidOutput\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minput1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     39\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 40\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mAlgoResponse\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcreate_algo_response\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mclient\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpostJsonHelper\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0murl\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0minput1\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mquery_parameters\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     41\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     42\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_postRawOutput\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0minput1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\ProgramData\\Anaconda3\\lib\\site-packages\\Algorithmia\\algo_response.py\u001b[0m in \u001b[0;36mcreate_algo_response\u001b[1;34m(responseJson)\u001b[0m\n\u001b[0;32m     18\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;34m'error'\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mresponseJson\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     19\u001b[0m             \u001b[1;31m# Failure\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 20\u001b[1;33m             \u001b[1;32mraise\u001b[0m \u001b[0mparse_exception\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mresponseJson\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'error'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     21\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     22\u001b[0m             \u001b[0mmetadata\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mMetadata\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mresponseJson\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'metadata'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mAlgorithmException\u001b[0m: 'Failed to run algorithm - may have timed-out or hit out-of-memory error'"
     ]
    }
   ],
   "source": [
    "import Algorithmia\n",
    "\n",
    "# Cambio el blend a 40 (-B 40) ya que fue el valor que mejores resultados dio en el paper\n",
    "input = {\n",
    "  \"trainUrl\": \"data://ludecan/EEG_Eye_State/EEGEyeState.arff\",\n",
    "  \"cv\": 3,\n",
    "  \"options\": \"-B 40\",\n",
    "  \"mode\": \"train\",\n",
    "  \"modelUrl\": \"data://ludecan/EEG_Eye_State/model.txt\"\n",
    "}\n",
    "client = Algorithmia.client('simqPUoVE90RoKHLf+4idN9o5Sc1')\n",
    "algo = client.algo('weka/KStar/0.1.1')\n",
    "algo = algo.set_options(timeout=3000)\n",
    "\n",
    "print(algo.pipe(input).result)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Clasificador Binario con Deep Learning\n",
    "Escalamiento de los datos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import preprocessing\n",
    "\n",
    "# Escalo los conjuntos de datos usando la media y escala del conjunto de entrenamiento\n",
    "scaler = preprocessing.StandardScaler().fit(train_features)\n",
    "sc_train_features = scaler.transform(train_features)\n",
    "sc_test_features = scaler.transform(test_features)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Creo un MLP con dos capas ocultas de 64 unidades cada una. Activación relu y agrego capas de dropout para evitar overfitting.\n",
    "Como medida de precisión vamos a usar accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "\n",
    "# Limpio la sesion de TensorFlow para empezar con un modelo de 0\n",
    "tf.keras.backend.clear_session()\n",
    "dl = keras.Sequential()\n",
    "\n",
    "dl.add(keras.layers.Dense(64, activation=tf.nn.relu, input_shape=(14,)))\n",
    "dl.add(keras.layers.Dropout(0.5))\n",
    "dl.add(keras.layers.Dense(64, activation=tf.nn.relu))\n",
    "dl.add(keras.layers.Dropout(0.5))\n",
    "dl.add(keras.layers.Dense(1, activation=tf.nn.sigmoid))\n",
    "\n",
    "dl.compile(optimizer='SGD',\n",
    "           loss='binary_crossentropy',\n",
    "           metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Entrenamiento del modelo. Le pongo 500 epochs como para que pueda evolucionar pero a su vez le pongo un callback de early stopping, para que no siga avanzando si\n",
    "no hay mejoras en la función de pérdida.\n",
    "Le pongo verbose = 0 para que no quede demasiado texto hacia abajo, si se quiere seguir el avance del entrenamiento poner verbose =1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "history = dl.fit(sc_train_features,\n",
    "                 train_labels,\n",
    "                 epochs=500,\n",
    "                 batch_size=128,\n",
    "                 callbacks=[keras.callbacks.EarlyStopping(monitor='val_loss',\n",
    "                                                          min_delta=0,\n",
    "                                                          patience=3,\n",
    "                                                          verbose=0, mode='auto')],\n",
    "                 validation_data=(sc_test_features, test_labels),\n",
    "                 verbose=0)"
   ]
  },
  {
   "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": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# gráfico de accuracy\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()\n",
    "# gráfico de función de pérdida\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()\n",
    "predictions_dl = dl.predict_classes(sc_test_features)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Clasificador Binario con Bosques Aleatorios\n",
    "Creo un clasificador de random forest y lo entreno con los datos de entrenamiento"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "rf = RandomForestClassifier(n_estimators= 100, random_state=rng, n_jobs=-1)\n",
    "rf.fit(train_features, train_labels);\n",
    "\n",
    "predictions_rf = rf.predict(test_features)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Importancia de variables"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Obtengo la importancia de las variables según el modelo ajustado\n",
    "En este caso O1, P7, F7, AF3, F8 y AF4 parecen ser las más significativas y las demás un poco menos.\n",
    "No lo hago ahora pero se podría plantear un clasificador con las variables más importantes y ver como se comportan las métricas de precisión"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable: O1                   Importancia: 0.12\n",
      "Variable: P7                   Importancia: 0.1\n",
      "Variable: F7                   Importancia: 0.09\n",
      "Variable: AF3                  Importancia: 0.08\n",
      "Variable: F8                   Importancia: 0.08\n",
      "Variable: AF4                  Importancia: 0.08\n",
      "Variable: FC5                  Importancia: 0.06\n",
      "Variable: O2                   Importancia: 0.06\n",
      "Variable: T8                   Importancia: 0.06\n",
      "Variable: FC6                  Importancia: 0.06\n",
      "Variable: F4                   Importancia: 0.06\n",
      "Variable: F3                   Importancia: 0.05\n",
      "Variable: T7                   Importancia: 0.05\n",
      "Variable: P8                   Importancia: 0.05\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "importances = list(rf.feature_importances_)\n",
    "feature_importances = [(feature, round(importance, 2)) for feature, importance in zip(feature_list, importances)]\n",
    "feature_importances = sorted(feature_importances, key = lambda x: x[1], reverse = True)\n",
    "[print('Variable: {:20} Importancia: {}'.format(*pair)) for pair in feature_importances];\n",
    "\n",
    "x_values = list(range(len(importances)))\n",
    "plt.bar(x_values, importances, orientation = 'vertical')\n",
    "plt.xticks(x_values, feature_list, rotation='vertical')\n",
    "plt.ylabel('Importance'); plt.xlabel('Variable'); plt.title('Variable Importances'); "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluación de los modelos\n",
    "Calculamos para cada modelo su accuracy y matriz de confusión"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluación de Deep Learning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8947262883497092"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "predictions_dl = np.transpose(predictions_dl)[0]\n",
    "\n",
    "accuracy_score(test_labels, predictions_dl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>Predicciones</th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Observaciones</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2582</td>\n",
       "      <td>221</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>304</td>\n",
       "      <td>1880</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Predicciones      0     1\n",
       "Observaciones            \n",
       "0              2582   221\n",
       "1               304  1880"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.crosstab(test_labels, predictions_dl, rownames=['Observaciones'], colnames=['Predicciones'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluación de Bosques Aleatorios"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9304190896330459"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(test_labels, predictions_rf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>Predicciones</th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Observaciones</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2672</td>\n",
       "      <td>131</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>216</td>\n",
       "      <td>1968</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Predicciones      0     1\n",
       "Observaciones            \n",
       "0              2672   131\n",
       "1               216  1968"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.crosstab(test_labels, predictions_rf, rownames=['Observaciones'], colnames=['Predicciones'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conclusión\n",
    "De los algoritmos aplicados que si funcionaron, random forest (RF) obtuvo una accuracy un poco mayor (0.93 vs 0.89) y mejores resultados en todas las celdas de la matriz de confusión por lo que para este caso resulta ganador. Sería interesante probar variantes del RF, como la versión con las variables más importantes, redes con arquitecturas más complejas e incluso hacer funcionar el K* y volver a comparar los resultados."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}