2.2-balancing.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Técnicas de balanceo de clases"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import sklearn.preprocessing\n",
    "import sklearn.linear_model\n",
    "import sklearn.svm\n",
    "import matplotlib.pyplot as plt\n",
    "import imblearn # Bilioteca para aprendizaje desde datos desbalanceados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "El problema del desbalance de clases es que la mayoría de las técnicas de aprendizaje de máquinas podrían subrepresentar y hasta ignorar la clase minoritaria, la cual casi siempre es la más importante en nuestra tarea de clasificación y a la cual enfocamos nuestros esfuerzos para mejorar su rendimiento."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preparación de los datos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "breast_cancer = pd.read_csv(\"data/breast_cancer.csv\") # Importamos un conjunto de datos de morfologías celulares y su asociación a malignidad."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>diagnosis</th>\n",
       "      <th>radius_mean</th>\n",
       "      <th>texture_mean</th>\n",
       "      <th>perimeter_mean</th>\n",
       "      <th>area_mean</th>\n",
       "      <th>smoothness_mean</th>\n",
       "      <th>compactness_mean</th>\n",
       "      <th>concavity_mean</th>\n",
       "      <th>concave points_mean</th>\n",
       "      <th>symmetry_mean</th>\n",
       "      <th>...</th>\n",
       "      <th>radius_worst</th>\n",
       "      <th>texture_worst</th>\n",
       "      <th>perimeter_worst</th>\n",
       "      <th>area_worst</th>\n",
       "      <th>smoothness_worst</th>\n",
       "      <th>compactness_worst</th>\n",
       "      <th>concavity_worst</th>\n",
       "      <th>concave points_worst</th>\n",
       "      <th>symmetry_worst</th>\n",
       "      <th>fractal_dimension_worst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>B</td>\n",
       "      <td>13.540</td>\n",
       "      <td>14.36</td>\n",
       "      <td>87.46</td>\n",
       "      <td>566.3</td>\n",
       "      <td>9.779</td>\n",
       "      <td>8.129</td>\n",
       "      <td>6.664</td>\n",
       "      <td>4.781</td>\n",
       "      <td>1.885</td>\n",
       "      <td>...</td>\n",
       "      <td>15.110</td>\n",
       "      <td>19.26</td>\n",
       "      <td>99.70</td>\n",
       "      <td>711.2</td>\n",
       "      <td>144.000</td>\n",
       "      <td>1.773</td>\n",
       "      <td>239.000</td>\n",
       "      <td>1.288</td>\n",
       "      <td>2.977</td>\n",
       "      <td>7.259</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>B</td>\n",
       "      <td>13.080</td>\n",
       "      <td>15.71</td>\n",
       "      <td>85.63</td>\n",
       "      <td>520.0</td>\n",
       "      <td>1.075</td>\n",
       "      <td>127.000</td>\n",
       "      <td>4.568</td>\n",
       "      <td>311.000</td>\n",
       "      <td>1.967</td>\n",
       "      <td>...</td>\n",
       "      <td>14.500</td>\n",
       "      <td>20.49</td>\n",
       "      <td>96.09</td>\n",
       "      <td>630.5</td>\n",
       "      <td>1.312</td>\n",
       "      <td>2.776</td>\n",
       "      <td>189.000</td>\n",
       "      <td>7.283</td>\n",
       "      <td>3.184</td>\n",
       "      <td>8.183</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>B</td>\n",
       "      <td>9.504</td>\n",
       "      <td>12.44</td>\n",
       "      <td>60.34</td>\n",
       "      <td>273.9</td>\n",
       "      <td>1.024</td>\n",
       "      <td>6.492</td>\n",
       "      <td>2.956</td>\n",
       "      <td>2.076</td>\n",
       "      <td>1.815</td>\n",
       "      <td>...</td>\n",
       "      <td>10.230</td>\n",
       "      <td>15.66</td>\n",
       "      <td>65.13</td>\n",
       "      <td>314.9</td>\n",
       "      <td>1.324</td>\n",
       "      <td>1.148</td>\n",
       "      <td>8.867</td>\n",
       "      <td>6.227</td>\n",
       "      <td>245.000</td>\n",
       "      <td>7.773</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>B</td>\n",
       "      <td>13.030</td>\n",
       "      <td>18.42</td>\n",
       "      <td>82.61</td>\n",
       "      <td>523.8</td>\n",
       "      <td>8.983</td>\n",
       "      <td>3.766</td>\n",
       "      <td>2.562</td>\n",
       "      <td>2.923</td>\n",
       "      <td>1.467</td>\n",
       "      <td>...</td>\n",
       "      <td>13.300</td>\n",
       "      <td>22.81</td>\n",
       "      <td>84.46</td>\n",
       "      <td>545.9</td>\n",
       "      <td>9.701</td>\n",
       "      <td>4.619</td>\n",
       "      <td>4.833</td>\n",
       "      <td>5.013</td>\n",
       "      <td>1.987</td>\n",
       "      <td>6.169</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>B</td>\n",
       "      <td>8.196</td>\n",
       "      <td>16.84</td>\n",
       "      <td>51.71</td>\n",
       "      <td>201.9</td>\n",
       "      <td>86.000</td>\n",
       "      <td>5.943</td>\n",
       "      <td>1.588</td>\n",
       "      <td>5.917</td>\n",
       "      <td>1.769</td>\n",
       "      <td>...</td>\n",
       "      <td>8.964</td>\n",
       "      <td>21.96</td>\n",
       "      <td>57.26</td>\n",
       "      <td>242.2</td>\n",
       "      <td>1.297</td>\n",
       "      <td>1.357</td>\n",
       "      <td>688.000</td>\n",
       "      <td>2.564</td>\n",
       "      <td>3.105</td>\n",
       "      <td>7.409</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 31 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "  diagnosis  radius_mean  texture_mean  perimeter_mean  area_mean  \\\n",
       "0         B       13.540         14.36           87.46      566.3   \n",
       "1         B       13.080         15.71           85.63      520.0   \n",
       "2         B        9.504         12.44           60.34      273.9   \n",
       "3         B       13.030         18.42           82.61      523.8   \n",
       "4         B        8.196         16.84           51.71      201.9   \n",
       "\n",
       "   smoothness_mean  compactness_mean  concavity_mean  concave points_mean  \\\n",
       "0            9.779             8.129           6.664                4.781   \n",
       "1            1.075           127.000           4.568              311.000   \n",
       "2            1.024             6.492           2.956                2.076   \n",
       "3            8.983             3.766           2.562                2.923   \n",
       "4           86.000             5.943           1.588                5.917   \n",
       "\n",
       "   symmetry_mean           ...             radius_worst  texture_worst  \\\n",
       "0          1.885           ...                   15.110          19.26   \n",
       "1          1.967           ...                   14.500          20.49   \n",
       "2          1.815           ...                   10.230          15.66   \n",
       "3          1.467           ...                   13.300          22.81   \n",
       "4          1.769           ...                    8.964          21.96   \n",
       "\n",
       "   perimeter_worst  area_worst  smoothness_worst  compactness_worst  \\\n",
       "0            99.70       711.2           144.000              1.773   \n",
       "1            96.09       630.5             1.312              2.776   \n",
       "2            65.13       314.9             1.324              1.148   \n",
       "3            84.46       545.9             9.701              4.619   \n",
       "4            57.26       242.2             1.297              1.357   \n",
       "\n",
       "   concavity_worst  concave points_worst  symmetry_worst  \\\n",
       "0          239.000                 1.288           2.977   \n",
       "1          189.000                 7.283           3.184   \n",
       "2            8.867                 6.227         245.000   \n",
       "3            4.833                 5.013           1.987   \n",
       "4          688.000                 2.564           3.105   \n",
       "\n",
       "   fractal_dimension_worst  \n",
       "0                    7.259  \n",
       "1                    8.183  \n",
       "2                    7.773  \n",
       "3                    6.169  \n",
       "4                    7.409  \n",
       "\n",
       "[5 rows x 31 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "breast_cancer.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "La columna diagnosis incluye la etiqueta asociada a esa observación."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "B    0.922481\n",
       "M    0.077519\n",
       "Name: diagnosis, dtype: float64"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "breast_cancer.diagnosis.value_counts( # Exploremos el balance de las clases del conjunto de dato\n",
    "    normalize = True # Normalizamos el resultado para obtener la proporción de las clases.\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sólo el 8 % de las observaciones pertenece a la clase M (Maligno)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = breast_cancer.iloc[:,1:] # Seleccionamos sólo las características y las guardamos en una variable.\n",
    "feature_names = features.columns # Guardamos los nombres de las características\n",
    "label = breast_cancer.diagnosis # La columna diagnosis contiene la etiqueta de los datos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['radius_mean', 'texture_mean', 'perimeter_mean', 'area_mean',\n",
       "       'smoothness_mean', 'compactness_mean', 'concavity_mean',\n",
       "       'concave points_mean', 'symmetry_mean', 'fractal_dimension_mean',\n",
       "       'radius_se', 'texture_se', 'perimeter_se', 'area_se', 'smoothness_se',\n",
       "       'compactness_se', 'concavity_se', 'concave points_se', 'symmetry_se',\n",
       "       'fractal_dimension_se', 'radius_worst', 'texture_worst',\n",
       "       'perimeter_worst', 'area_worst', 'smoothness_worst',\n",
       "       'compactness_worst', 'concavity_worst', 'concave points_worst',\n",
       "       'symmetry_worst', 'fractal_dimension_worst'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_names"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Primero verificamos los rangos de las variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>radius_mean</th>\n",
       "      <th>texture_mean</th>\n",
       "      <th>perimeter_mean</th>\n",
       "      <th>area_mean</th>\n",
       "      <th>smoothness_mean</th>\n",
       "      <th>compactness_mean</th>\n",
       "      <th>concavity_mean</th>\n",
       "      <th>concave points_mean</th>\n",
       "      <th>symmetry_mean</th>\n",
       "      <th>fractal_dimension_mean</th>\n",
       "      <th>...</th>\n",
       "      <th>radius_worst</th>\n",
       "      <th>texture_worst</th>\n",
       "      <th>perimeter_worst</th>\n",
       "      <th>area_worst</th>\n",
       "      <th>smoothness_worst</th>\n",
       "      <th>compactness_worst</th>\n",
       "      <th>concavity_worst</th>\n",
       "      <th>concave points_worst</th>\n",
       "      <th>symmetry_worst</th>\n",
       "      <th>fractal_dimension_worst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>12.469481</td>\n",
       "      <td>18.113747</td>\n",
       "      <td>80.410465</td>\n",
       "      <td>492.932041</td>\n",
       "      <td>49.744455</td>\n",
       "      <td>35.652729</td>\n",
       "      <td>43.771661</td>\n",
       "      <td>32.595760</td>\n",
       "      <td>21.960455</td>\n",
       "      <td>64.511034</td>\n",
       "      <td>...</td>\n",
       "      <td>13.920488</td>\n",
       "      <td>23.914651</td>\n",
       "      <td>90.925168</td>\n",
       "      <td>618.007752</td>\n",
       "      <td>13.864899</td>\n",
       "      <td>33.948514</td>\n",
       "      <td>39.414204</td>\n",
       "      <td>54.998109</td>\n",
       "      <td>27.036059</td>\n",
       "      <td>59.537822</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>2.176057</td>\n",
       "      <td>4.041660</td>\n",
       "      <td>14.754203</td>\n",
       "      <td>183.335352</td>\n",
       "      <td>187.941745</td>\n",
       "      <td>135.979181</td>\n",
       "      <td>150.341273</td>\n",
       "      <td>104.720222</td>\n",
       "      <td>58.314459</td>\n",
       "      <td>182.662695</td>\n",
       "      <td>...</td>\n",
       "      <td>2.871213</td>\n",
       "      <td>5.693853</td>\n",
       "      <td>20.076597</td>\n",
       "      <td>295.328421</td>\n",
       "      <td>60.586437</td>\n",
       "      <td>115.481235</td>\n",
       "      <td>133.193922</td>\n",
       "      <td>179.546410</td>\n",
       "      <td>78.141293</td>\n",
       "      <td>195.397602</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>6.981000</td>\n",
       "      <td>9.710000</td>\n",
       "      <td>43.790000</td>\n",
       "      <td>143.500000</td>\n",
       "      <td>0.110000</td>\n",
       "      <td>0.060000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.180000</td>\n",
       "      <td>5.185000</td>\n",
       "      <td>...</td>\n",
       "      <td>7.930000</td>\n",
       "      <td>12.020000</td>\n",
       "      <td>50.410000</td>\n",
       "      <td>185.200000</td>\n",
       "      <td>0.130000</td>\n",
       "      <td>0.100000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.210000</td>\n",
       "      <td>0.120000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>11.220000</td>\n",
       "      <td>15.195000</td>\n",
       "      <td>71.490000</td>\n",
       "      <td>386.050000</td>\n",
       "      <td>1.230500</td>\n",
       "      <td>1.997000</td>\n",
       "      <td>2.038000</td>\n",
       "      <td>1.931500</td>\n",
       "      <td>1.597000</td>\n",
       "      <td>5.883000</td>\n",
       "      <td>...</td>\n",
       "      <td>12.320000</td>\n",
       "      <td>19.655000</td>\n",
       "      <td>79.375000</td>\n",
       "      <td>458.000000</td>\n",
       "      <td>1.202500</td>\n",
       "      <td>1.665500</td>\n",
       "      <td>1.606500</td>\n",
       "      <td>1.617000</td>\n",
       "      <td>2.445500</td>\n",
       "      <td>6.774500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>12.340000</td>\n",
       "      <td>17.570000</td>\n",
       "      <td>78.850000</td>\n",
       "      <td>466.100000</td>\n",
       "      <td>8.306000</td>\n",
       "      <td>5.688000</td>\n",
       "      <td>3.735000</td>\n",
       "      <td>2.924000</td>\n",
       "      <td>1.769000</td>\n",
       "      <td>6.246000</td>\n",
       "      <td>...</td>\n",
       "      <td>13.580000</td>\n",
       "      <td>23.070000</td>\n",
       "      <td>88.100000</td>\n",
       "      <td>562.600000</td>\n",
       "      <td>1.338000</td>\n",
       "      <td>2.399000</td>\n",
       "      <td>2.671000</td>\n",
       "      <td>5.602000</td>\n",
       "      <td>2.768000</td>\n",
       "      <td>7.628000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>13.645000</td>\n",
       "      <td>20.200000</td>\n",
       "      <td>87.860000</td>\n",
       "      <td>572.050000</td>\n",
       "      <td>9.264000</td>\n",
       "      <td>7.815500</td>\n",
       "      <td>6.539500</td>\n",
       "      <td>5.548000</td>\n",
       "      <td>1.998000</td>\n",
       "      <td>6.907500</td>\n",
       "      <td>...</td>\n",
       "      <td>15.105000</td>\n",
       "      <td>27.260000</td>\n",
       "      <td>98.955000</td>\n",
       "      <td>700.650000</td>\n",
       "      <td>1.546000</td>\n",
       "      <td>5.805500</td>\n",
       "      <td>6.237000</td>\n",
       "      <td>8.249000</td>\n",
       "      <td>3.159000</td>\n",
       "      <td>8.472500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>21.160000</td>\n",
       "      <td>33.810000</td>\n",
       "      <td>137.200000</td>\n",
       "      <td>1404.000000</td>\n",
       "      <td>997.000000</td>\n",
       "      <td>958.000000</td>\n",
       "      <td>973.000000</td>\n",
       "      <td>917.000000</td>\n",
       "      <td>304.000000</td>\n",
       "      <td>898.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>29.170000</td>\n",
       "      <td>41.780000</td>\n",
       "      <td>188.000000</td>\n",
       "      <td>2615.000000</td>\n",
       "      <td>997.000000</td>\n",
       "      <td>937.000000</td>\n",
       "      <td>942.000000</td>\n",
       "      <td>991.000000</td>\n",
       "      <td>353.000000</td>\n",
       "      <td>997.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       radius_mean  texture_mean  perimeter_mean    area_mean  \\\n",
       "count   387.000000    387.000000      387.000000   387.000000   \n",
       "mean     12.469481     18.113747       80.410465   492.932041   \n",
       "std       2.176057      4.041660       14.754203   183.335352   \n",
       "min       6.981000      9.710000       43.790000   143.500000   \n",
       "25%      11.220000     15.195000       71.490000   386.050000   \n",
       "50%      12.340000     17.570000       78.850000   466.100000   \n",
       "75%      13.645000     20.200000       87.860000   572.050000   \n",
       "max      21.160000     33.810000      137.200000  1404.000000   \n",
       "\n",
       "       smoothness_mean  compactness_mean  concavity_mean  concave points_mean  \\\n",
       "count       387.000000        387.000000      387.000000           387.000000   \n",
       "mean         49.744455         35.652729       43.771661            32.595760   \n",
       "std         187.941745        135.979181      150.341273           104.720222   \n",
       "min           0.110000          0.060000        0.000000             0.000000   \n",
       "25%           1.230500          1.997000        2.038000             1.931500   \n",
       "50%           8.306000          5.688000        3.735000             2.924000   \n",
       "75%           9.264000          7.815500        6.539500             5.548000   \n",
       "max         997.000000        958.000000      973.000000           917.000000   \n",
       "\n",
       "       symmetry_mean  fractal_dimension_mean           ...             \\\n",
       "count     387.000000              387.000000           ...              \n",
       "mean       21.960455               64.511034           ...              \n",
       "std        58.314459              182.662695           ...              \n",
       "min         0.180000                5.185000           ...              \n",
       "25%         1.597000                5.883000           ...              \n",
       "50%         1.769000                6.246000           ...              \n",
       "75%         1.998000                6.907500           ...              \n",
       "max       304.000000              898.000000           ...              \n",
       "\n",
       "       radius_worst  texture_worst  perimeter_worst   area_worst  \\\n",
       "count    387.000000     387.000000       387.000000   387.000000   \n",
       "mean      13.920488      23.914651        90.925168   618.007752   \n",
       "std        2.871213       5.693853        20.076597   295.328421   \n",
       "min        7.930000      12.020000        50.410000   185.200000   \n",
       "25%       12.320000      19.655000        79.375000   458.000000   \n",
       "50%       13.580000      23.070000        88.100000   562.600000   \n",
       "75%       15.105000      27.260000        98.955000   700.650000   \n",
       "max       29.170000      41.780000       188.000000  2615.000000   \n",
       "\n",
       "       smoothness_worst  compactness_worst  concavity_worst  \\\n",
       "count        387.000000         387.000000       387.000000   \n",
       "mean          13.864899          33.948514        39.414204   \n",
       "std           60.586437         115.481235       133.193922   \n",
       "min            0.130000           0.100000         0.000000   \n",
       "25%            1.202500           1.665500         1.606500   \n",
       "50%            1.338000           2.399000         2.671000   \n",
       "75%            1.546000           5.805500         6.237000   \n",
       "max          997.000000         937.000000       942.000000   \n",
       "\n",
       "       concave points_worst  symmetry_worst  fractal_dimension_worst  \n",
       "count            387.000000      387.000000               387.000000  \n",
       "mean              54.998109       27.036059                59.537822  \n",
       "std              179.546410       78.141293               195.397602  \n",
       "min                0.000000        0.210000                 0.120000  \n",
       "25%                1.617000        2.445500                 6.774500  \n",
       "50%                5.602000        2.768000                 7.628000  \n",
       "75%                8.249000        3.159000                 8.472500  \n",
       "max              991.000000      353.000000               997.000000  \n",
       "\n",
       "[8 rows x 30 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Procedemos a escalar los datos porque las características tienen distintos rangos."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "scaler = sklearn.preprocessing.MinMaxScaler() \n",
    "features = pd.DataFrame(scaler.fit_transform(features), columns=feature_names)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ahora todas las características se encuentran en el rango entre 0 y 1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>radius_mean</th>\n",
       "      <th>texture_mean</th>\n",
       "      <th>perimeter_mean</th>\n",
       "      <th>area_mean</th>\n",
       "      <th>smoothness_mean</th>\n",
       "      <th>compactness_mean</th>\n",
       "      <th>concavity_mean</th>\n",
       "      <th>concave points_mean</th>\n",
       "      <th>symmetry_mean</th>\n",
       "      <th>fractal_dimension_mean</th>\n",
       "      <th>...</th>\n",
       "      <th>radius_worst</th>\n",
       "      <th>texture_worst</th>\n",
       "      <th>perimeter_worst</th>\n",
       "      <th>area_worst</th>\n",
       "      <th>smoothness_worst</th>\n",
       "      <th>compactness_worst</th>\n",
       "      <th>concavity_worst</th>\n",
       "      <th>concave points_worst</th>\n",
       "      <th>symmetry_worst</th>\n",
       "      <th>fractal_dimension_worst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "      <td>387.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.387085</td>\n",
       "      <td>0.348703</td>\n",
       "      <td>0.392040</td>\n",
       "      <td>0.277217</td>\n",
       "      <td>0.049789</td>\n",
       "      <td>0.037155</td>\n",
       "      <td>0.044986</td>\n",
       "      <td>0.035546</td>\n",
       "      <td>0.071689</td>\n",
       "      <td>0.066448</td>\n",
       "      <td>...</td>\n",
       "      <td>0.282038</td>\n",
       "      <td>0.399686</td>\n",
       "      <td>0.294463</td>\n",
       "      <td>0.178125</td>\n",
       "      <td>0.013778</td>\n",
       "      <td>0.036128</td>\n",
       "      <td>0.041841</td>\n",
       "      <td>0.055498</td>\n",
       "      <td>0.076040</td>\n",
       "      <td>0.059604</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.153470</td>\n",
       "      <td>0.167704</td>\n",
       "      <td>0.157951</td>\n",
       "      <td>0.145447</td>\n",
       "      <td>0.188528</td>\n",
       "      <td>0.141950</td>\n",
       "      <td>0.154513</td>\n",
       "      <td>0.114199</td>\n",
       "      <td>0.191938</td>\n",
       "      <td>0.204592</td>\n",
       "      <td>...</td>\n",
       "      <td>0.135180</td>\n",
       "      <td>0.191326</td>\n",
       "      <td>0.145916</td>\n",
       "      <td>0.121544</td>\n",
       "      <td>0.060777</td>\n",
       "      <td>0.123259</td>\n",
       "      <td>0.141395</td>\n",
       "      <td>0.181177</td>\n",
       "      <td>0.221495</td>\n",
       "      <td>0.196009</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.298963</td>\n",
       "      <td>0.227593</td>\n",
       "      <td>0.296542</td>\n",
       "      <td>0.192424</td>\n",
       "      <td>0.001124</td>\n",
       "      <td>0.002022</td>\n",
       "      <td>0.002095</td>\n",
       "      <td>0.002106</td>\n",
       "      <td>0.004664</td>\n",
       "      <td>0.000782</td>\n",
       "      <td>...</td>\n",
       "      <td>0.206685</td>\n",
       "      <td>0.256552</td>\n",
       "      <td>0.210517</td>\n",
       "      <td>0.112273</td>\n",
       "      <td>0.001076</td>\n",
       "      <td>0.001671</td>\n",
       "      <td>0.001705</td>\n",
       "      <td>0.001632</td>\n",
       "      <td>0.006337</td>\n",
       "      <td>0.006675</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.377953</td>\n",
       "      <td>0.326141</td>\n",
       "      <td>0.375335</td>\n",
       "      <td>0.255930</td>\n",
       "      <td>0.008222</td>\n",
       "      <td>0.005875</td>\n",
       "      <td>0.003839</td>\n",
       "      <td>0.003189</td>\n",
       "      <td>0.005230</td>\n",
       "      <td>0.001188</td>\n",
       "      <td>...</td>\n",
       "      <td>0.266008</td>\n",
       "      <td>0.371304</td>\n",
       "      <td>0.273930</td>\n",
       "      <td>0.155321</td>\n",
       "      <td>0.001212</td>\n",
       "      <td>0.002454</td>\n",
       "      <td>0.002835</td>\n",
       "      <td>0.005653</td>\n",
       "      <td>0.007251</td>\n",
       "      <td>0.007531</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.469991</td>\n",
       "      <td>0.435270</td>\n",
       "      <td>0.471791</td>\n",
       "      <td>0.339984</td>\n",
       "      <td>0.009183</td>\n",
       "      <td>0.008096</td>\n",
       "      <td>0.006721</td>\n",
       "      <td>0.006050</td>\n",
       "      <td>0.005984</td>\n",
       "      <td>0.001929</td>\n",
       "      <td>...</td>\n",
       "      <td>0.337806</td>\n",
       "      <td>0.512097</td>\n",
       "      <td>0.352824</td>\n",
       "      <td>0.212137</td>\n",
       "      <td>0.001420</td>\n",
       "      <td>0.006090</td>\n",
       "      <td>0.006621</td>\n",
       "      <td>0.008324</td>\n",
       "      <td>0.008359</td>\n",
       "      <td>0.008379</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       radius_mean  texture_mean  perimeter_mean   area_mean  smoothness_mean  \\\n",
       "count   387.000000    387.000000      387.000000  387.000000       387.000000   \n",
       "mean      0.387085      0.348703        0.392040    0.277217         0.049789   \n",
       "std       0.153470      0.167704        0.157951    0.145447         0.188528   \n",
       "min       0.000000      0.000000        0.000000    0.000000         0.000000   \n",
       "25%       0.298963      0.227593        0.296542    0.192424         0.001124   \n",
       "50%       0.377953      0.326141        0.375335    0.255930         0.008222   \n",
       "75%       0.469991      0.435270        0.471791    0.339984         0.009183   \n",
       "max       1.000000      1.000000        1.000000    1.000000         1.000000   \n",
       "\n",
       "       compactness_mean  concavity_mean  concave points_mean  symmetry_mean  \\\n",
       "count        387.000000      387.000000           387.000000     387.000000   \n",
       "mean           0.037155        0.044986             0.035546       0.071689   \n",
       "std            0.141950        0.154513             0.114199       0.191938   \n",
       "min            0.000000        0.000000             0.000000       0.000000   \n",
       "25%            0.002022        0.002095             0.002106       0.004664   \n",
       "50%            0.005875        0.003839             0.003189       0.005230   \n",
       "75%            0.008096        0.006721             0.006050       0.005984   \n",
       "max            1.000000        1.000000             1.000000       1.000000   \n",
       "\n",
       "       fractal_dimension_mean           ...             radius_worst  \\\n",
       "count              387.000000           ...               387.000000   \n",
       "mean                 0.066448           ...                 0.282038   \n",
       "std                  0.204592           ...                 0.135180   \n",
       "min                  0.000000           ...                 0.000000   \n",
       "25%                  0.000782           ...                 0.206685   \n",
       "50%                  0.001188           ...                 0.266008   \n",
       "75%                  0.001929           ...                 0.337806   \n",
       "max                  1.000000           ...                 1.000000   \n",
       "\n",
       "       texture_worst  perimeter_worst  area_worst  smoothness_worst  \\\n",
       "count     387.000000       387.000000  387.000000        387.000000   \n",
       "mean        0.399686         0.294463    0.178125          0.013778   \n",
       "std         0.191326         0.145916    0.121544          0.060777   \n",
       "min         0.000000         0.000000    0.000000          0.000000   \n",
       "25%         0.256552         0.210517    0.112273          0.001076   \n",
       "50%         0.371304         0.273930    0.155321          0.001212   \n",
       "75%         0.512097         0.352824    0.212137          0.001420   \n",
       "max         1.000000         1.000000    1.000000          1.000000   \n",
       "\n",
       "       compactness_worst  concavity_worst  concave points_worst  \\\n",
       "count         387.000000       387.000000            387.000000   \n",
       "mean            0.036128         0.041841              0.055498   \n",
       "std             0.123259         0.141395              0.181177   \n",
       "min             0.000000         0.000000              0.000000   \n",
       "25%             0.001671         0.001705              0.001632   \n",
       "50%             0.002454         0.002835              0.005653   \n",
       "75%             0.006090         0.006621              0.008324   \n",
       "max             1.000000         1.000000              1.000000   \n",
       "\n",
       "       symmetry_worst  fractal_dimension_worst  \n",
       "count      387.000000               387.000000  \n",
       "mean         0.076040                 0.059604  \n",
       "std          0.221495                 0.196009  \n",
       "min          0.000000                 0.000000  \n",
       "25%          0.006337                 0.006675  \n",
       "50%          0.007251                 0.007531  \n",
       "75%          0.008359                 0.008379  \n",
       "max          1.000000                 1.000000  \n",
       "\n",
       "[8 rows x 30 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "features_train, features_test, label_train, label_test = sklearn.model_selection.train_test_split(\n",
    "    features,\n",
    "    label,\n",
    "    test_size=0.30, \n",
    "    random_state = 11,\n",
    "    stratify = label\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para explorar los datos construiremos un gráfico de dispersión con un par de variables y la etiqueta correspondiente de la observación."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "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.scatter(features_train[label_train == \"B\"].area_worst, \n",
    "            features_train[label_train == \"B\"].perimeter_worst, \n",
    "            label=\"B\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.scatter(features_train[label_train == \"M\"].area_worst, \n",
    "            features_train[label_train == \"M\"].perimeter_worst, \n",
    "            label=\"M\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.title(\"Tissue malignancy by cell measurements\")\n",
    "plt.xlabel(\"area_worst\")\n",
    "plt.ylabel(\"perimeter_worst\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "En el gráfico anterior se puede observar claramente el desbalance de las clases, donde la clase M está sumamente subrepresentada en el conjunto de datos."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Observemos primero las medidas de tendencia central que tienen las características de la clase M."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>radius_mean</th>\n",
       "      <th>texture_mean</th>\n",
       "      <th>perimeter_mean</th>\n",
       "      <th>area_mean</th>\n",
       "      <th>smoothness_mean</th>\n",
       "      <th>compactness_mean</th>\n",
       "      <th>concavity_mean</th>\n",
       "      <th>concave points_mean</th>\n",
       "      <th>symmetry_mean</th>\n",
       "      <th>fractal_dimension_mean</th>\n",
       "      <th>...</th>\n",
       "      <th>radius_worst</th>\n",
       "      <th>texture_worst</th>\n",
       "      <th>perimeter_worst</th>\n",
       "      <th>area_worst</th>\n",
       "      <th>smoothness_worst</th>\n",
       "      <th>compactness_worst</th>\n",
       "      <th>concavity_worst</th>\n",
       "      <th>concave points_worst</th>\n",
       "      <th>symmetry_worst</th>\n",
       "      <th>fractal_dimension_worst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.663486</td>\n",
       "      <td>0.439123</td>\n",
       "      <td>0.692003</td>\n",
       "      <td>0.573061</td>\n",
       "      <td>0.141615</td>\n",
       "      <td>0.010819</td>\n",
       "      <td>0.062898</td>\n",
       "      <td>0.057599</td>\n",
       "      <td>0.073993</td>\n",
       "      <td>0.038804</td>\n",
       "      <td>...</td>\n",
       "      <td>0.580890</td>\n",
       "      <td>0.543859</td>\n",
       "      <td>0.622038</td>\n",
       "      <td>0.467213</td>\n",
       "      <td>0.007940</td>\n",
       "      <td>0.030168</td>\n",
       "      <td>0.031266</td>\n",
       "      <td>0.058382</td>\n",
       "      <td>0.046244</td>\n",
       "      <td>0.013335</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.217471</td>\n",
       "      <td>0.126019</td>\n",
       "      <td>0.212234</td>\n",
       "      <td>0.256124</td>\n",
       "      <td>0.344882</td>\n",
       "      <td>0.028494</td>\n",
       "      <td>0.197477</td>\n",
       "      <td>0.216554</td>\n",
       "      <td>0.215491</td>\n",
       "      <td>0.153229</td>\n",
       "      <td>...</td>\n",
       "      <td>0.201722</td>\n",
       "      <td>0.185848</td>\n",
       "      <td>0.199145</td>\n",
       "      <td>0.241883</td>\n",
       "      <td>0.030101</td>\n",
       "      <td>0.084298</td>\n",
       "      <td>0.123956</td>\n",
       "      <td>0.092937</td>\n",
       "      <td>0.167898</td>\n",
       "      <td>0.036844</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.313069</td>\n",
       "      <td>0.188797</td>\n",
       "      <td>0.361739</td>\n",
       "      <td>0.192463</td>\n",
       "      <td>0.000896</td>\n",
       "      <td>0.000983</td>\n",
       "      <td>0.001127</td>\n",
       "      <td>0.001137</td>\n",
       "      <td>0.004615</td>\n",
       "      <td>0.000104</td>\n",
       "      <td>...</td>\n",
       "      <td>0.328625</td>\n",
       "      <td>0.156250</td>\n",
       "      <td>0.343339</td>\n",
       "      <td>0.157420</td>\n",
       "      <td>0.000910</td>\n",
       "      <td>0.000171</td>\n",
       "      <td>0.000425</td>\n",
       "      <td>0.001129</td>\n",
       "      <td>0.006040</td>\n",
       "      <td>0.000906</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.475986</td>\n",
       "      <td>0.373029</td>\n",
       "      <td>0.533241</td>\n",
       "      <td>0.344942</td>\n",
       "      <td>0.001014</td>\n",
       "      <td>0.001286</td>\n",
       "      <td>0.001567</td>\n",
       "      <td>0.005735</td>\n",
       "      <td>0.005312</td>\n",
       "      <td>0.000624</td>\n",
       "      <td>...</td>\n",
       "      <td>0.419492</td>\n",
       "      <td>0.486559</td>\n",
       "      <td>0.447634</td>\n",
       "      <td>0.289530</td>\n",
       "      <td>0.001248</td>\n",
       "      <td>0.002153</td>\n",
       "      <td>0.002843</td>\n",
       "      <td>0.001640</td>\n",
       "      <td>0.007404</td>\n",
       "      <td>0.001286</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.625503</td>\n",
       "      <td>0.461411</td>\n",
       "      <td>0.641366</td>\n",
       "      <td>0.507100</td>\n",
       "      <td>0.001319</td>\n",
       "      <td>0.001655</td>\n",
       "      <td>0.002187</td>\n",
       "      <td>0.008431</td>\n",
       "      <td>0.005974</td>\n",
       "      <td>0.001080</td>\n",
       "      <td>...</td>\n",
       "      <td>0.568738</td>\n",
       "      <td>0.540995</td>\n",
       "      <td>0.614071</td>\n",
       "      <td>0.428760</td>\n",
       "      <td>0.001338</td>\n",
       "      <td>0.004059</td>\n",
       "      <td>0.004017</td>\n",
       "      <td>0.002228</td>\n",
       "      <td>0.009598</td>\n",
       "      <td>0.007429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.859652</td>\n",
       "      <td>0.516183</td>\n",
       "      <td>0.922920</td>\n",
       "      <td>0.777073</td>\n",
       "      <td>0.009751</td>\n",
       "      <td>0.002439</td>\n",
       "      <td>0.010149</td>\n",
       "      <td>0.009578</td>\n",
       "      <td>0.006981</td>\n",
       "      <td>0.002797</td>\n",
       "      <td>...</td>\n",
       "      <td>0.703861</td>\n",
       "      <td>0.633737</td>\n",
       "      <td>0.741987</td>\n",
       "      <td>0.584739</td>\n",
       "      <td>0.001529</td>\n",
       "      <td>0.006248</td>\n",
       "      <td>0.005880</td>\n",
       "      <td>0.150353</td>\n",
       "      <td>0.011780</td>\n",
       "      <td>0.008274</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.626141</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.976928</td>\n",
       "      <td>0.113723</td>\n",
       "      <td>0.893114</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.772892</td>\n",
       "      <td>0.703186</td>\n",
       "      <td>...</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.963038</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.139306</td>\n",
       "      <td>0.336108</td>\n",
       "      <td>0.572187</td>\n",
       "      <td>0.245207</td>\n",
       "      <td>0.778905</td>\n",
       "      <td>0.173421</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       radius_mean  texture_mean  perimeter_mean  area_mean  smoothness_mean  \\\n",
       "count    21.000000     21.000000       21.000000  21.000000        21.000000   \n",
       "mean      0.663486      0.439123        0.692003   0.573061         0.141615   \n",
       "std       0.217471      0.126019        0.212234   0.256124         0.344882   \n",
       "min       0.313069      0.188797        0.361739   0.192463         0.000896   \n",
       "25%       0.475986      0.373029        0.533241   0.344942         0.001014   \n",
       "50%       0.625503      0.461411        0.641366   0.507100         0.001319   \n",
       "75%       0.859652      0.516183        0.922920   0.777073         0.009751   \n",
       "max       1.000000      0.626141        1.000000   1.000000         0.976928   \n",
       "\n",
       "       compactness_mean  concavity_mean  concave points_mean  symmetry_mean  \\\n",
       "count         21.000000       21.000000            21.000000      21.000000   \n",
       "mean           0.010819        0.062898             0.057599       0.073993   \n",
       "std            0.028494        0.197477             0.216554       0.215491   \n",
       "min            0.000983        0.001127             0.001137       0.004615   \n",
       "25%            0.001286        0.001567             0.005735       0.005312   \n",
       "50%            0.001655        0.002187             0.008431       0.005974   \n",
       "75%            0.002439        0.010149             0.009578       0.006981   \n",
       "max            0.113723        0.893114             1.000000       0.772892   \n",
       "\n",
       "       fractal_dimension_mean           ...             radius_worst  \\\n",
       "count               21.000000           ...                21.000000   \n",
       "mean                 0.038804           ...                 0.580890   \n",
       "std                  0.153229           ...                 0.201722   \n",
       "min                  0.000104           ...                 0.328625   \n",
       "25%                  0.000624           ...                 0.419492   \n",
       "50%                  0.001080           ...                 0.568738   \n",
       "75%                  0.002797           ...                 0.703861   \n",
       "max                  0.703186           ...                 1.000000   \n",
       "\n",
       "       texture_worst  perimeter_worst  area_worst  smoothness_worst  \\\n",
       "count      21.000000        21.000000   21.000000         21.000000   \n",
       "mean        0.543859         0.622038    0.467213          0.007940   \n",
       "std         0.185848         0.199145    0.241883          0.030101   \n",
       "min         0.156250         0.343339    0.157420          0.000910   \n",
       "25%         0.486559         0.447634    0.289530          0.001248   \n",
       "50%         0.540995         0.614071    0.428760          0.001338   \n",
       "75%         0.633737         0.741987    0.584739          0.001529   \n",
       "max         0.963038         1.000000    1.000000          0.139306   \n",
       "\n",
       "       compactness_worst  concavity_worst  concave points_worst  \\\n",
       "count          21.000000        21.000000             21.000000   \n",
       "mean            0.030168         0.031266              0.058382   \n",
       "std             0.084298         0.123956              0.092937   \n",
       "min             0.000171         0.000425              0.001129   \n",
       "25%             0.002153         0.002843              0.001640   \n",
       "50%             0.004059         0.004017              0.002228   \n",
       "75%             0.006248         0.005880              0.150353   \n",
       "max             0.336108         0.572187              0.245207   \n",
       "\n",
       "       symmetry_worst  fractal_dimension_worst  \n",
       "count       21.000000                21.000000  \n",
       "mean         0.046244                 0.013335  \n",
       "std          0.167898                 0.036844  \n",
       "min          0.006040                 0.000906  \n",
       "25%          0.007404                 0.001286  \n",
       "50%          0.009598                 0.007429  \n",
       "75%          0.011780                 0.008274  \n",
       "max          0.778905                 0.173421  \n",
       "\n",
       "[8 rows x 30 columns]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features_train[label_train == \"M\"].describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Clases desbalanceadas"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Entrenemos un modelo con las clases desbalanceadas."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "                   intercept_scaling=1, l1_ratio=None, max_iter=100,\n",
       "                   multi_class='auto', n_jobs=None, penalty='l2',\n",
       "                   random_state=11, solver='lbfgs', tol=0.0001, verbose=0,\n",
       "                   warm_start=False)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr = sklearn.linear_model.LogisticRegression(random_state=11)\n",
    "lr.fit(features_train, label_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr_predictions = lr.predict(features_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           B       0.96      1.00      0.98       108\n",
      "           M       1.00      0.56      0.71         9\n",
      "\n",
      "    accuracy                           0.97       117\n",
      "   macro avg       0.98      0.78      0.85       117\n",
      "weighted avg       0.97      0.97      0.96       117\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(sklearn.metrics.classification_report(label_test, lr_predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "La métrica que salta a la luz es el bajo recall de la clase M."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Submuestreo de la clase mayoritaria"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "El enfoque más ingenuo para corregir el balance de los datos sería utilizar sólo un subconjunto de la clase sobrerepresentada que tenga el mismo tamaño que la clase subrepresentada.\n",
    "\n",
    "Los métodos utilizados para remuestrear las clases se encuentran en la biblioteca imblearn. La técnica enunciada anteriormente se encuentra en la clase imblearn.under_sampling.RandomUnderSampler(). Esta clase implementa métodos para submuestrar al azar la clase mayoritaria para igualar su frecuencia a la de la clase minoritaria."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "rus = imblearn.under_sampling.RandomUnderSampler(random_state=11) # Instanciamos la clase para submuestrear nuestro conjunto de datos.\n",
    "features_train_rus, label_train_rus = rus.fit_resample(features_train, label_train) # Con este método remuestreamos \n",
    "                                                                                    # las características y la etiqueta \n",
    "                                                                                    # y las guardamos en variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEXCAYAAACpuuMDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzt3Xuc3HV97/HXO7ub7JJsdnMjhNxVoIEIiOHiDVHEBo6EnqMVaAGt2rRarK21PVJPj4htUevlCNJT03rBG4habKA2SJVbRTwEBQwstDGJZEMISSCzCZkNe/mcP36/TWY2e5nJzm9nZ/f9fDzymPn95jczn9/OZD6/710RgZmZWZ9J1Q7AzMzGFicGMzMr4sRgZmZFnBjMzKyIE4OZmRVxYjAzsyJODFUk6S8l/VO146g0SedIai/YfkzSOVUMqWIkfVXSX1fx/ZdICkn16fbdkt5TrXhsfKqvdgDjmaR9BZtHAQeAnnT7DyLib0c/qtEXESdVOwazckm6G/hGRIy7i7fhODFkKCKm9d2XtAV4T0T8e/UiMhu7JNVHRHe14zBXJVWVpKslfSO93yjpG5J2S9oj6UFJc9PH3ilpk6S9kjZL+t3+z0+3+1cztEj6kqTtkrZJ+mtJdUPE8p00hr2SfinpeElXSXpW0lZJby44/vcktaXHbpL0B0Oc5xZJb0rvN0m6UdLz6fP/ol+10xZJH5L0qKScpG9LakwfmyHpdkk70+ffLmlBwXPvlvRxST9J4/qhpNkFj79W0v3p33dr+nc9XdKOvr9ZetxbJT08xEc3W9Kd6XvcI2lx+rwbJH2m37nfJulPBvm7nJS+znNpDH+Z7p8k6cOSfpV+H26RNHOIeAZ0BJ/poN8XSS+V9OM0nl2SvimpteC5/zN9zl5JT0o6N91fVPWmw6sZt6TPfRR4QVK9pGMlfS/9nDdL+uOMzumdkv5D0qfT79NmSeenj/0N8DrgC5L2SfqCEp9L3yeXfkeXl/u51AInhrHjHUALsBCYBfwhkJc0FbgOOD8imoFXA0P9aBW6EegGXga8AngzMFR99IXA14EZwC+AO0i+I/OBa4AvFhz7LPAWYDrwe8DnJJ1WQkwfBZYALwHOAy4b4Ji3AyuBpcDJwDvT/ZOArwCLgUVAHvhCv+f+ThrP0cBk4EMAkhYB/wZcD8wBTgUejogHgd1pLH0uS/8Og/ld4OPAbJLP4pvp/huBSyVNSt9zNnAucFP/F5DUDPw7sA44luQz+lH68B8DvwW8Pn3seeCGIeIZSjmf6VDfFwHXpvEsI/meXp2eywnAlcDp6Xf0N4EtZcR4KfDfgFagF7gNeCSN8VzgTyT9ZgbnBHAm8CTJZ/kp4EuSFBEfAe4DroyIaRFxZfrcs4Hj01gvJvnujD8R4X+j8I/kP8qb+u27mqQOE+BdwP3Ayf2OmQrsAd4KNA32/HR7CRAkVYRzSdo0mgoevxS4a5D4rgbuLNi+ENgH1KXbzelrtw7y/O8DH0jvnwO0D3TuwCbgNwsee88Ax15WsP0p4B8Gec9TgecLtu8G/lfB9vuAden9q4BbB3md/wl8M70/E9gPzBvk2K8CNxdsTyNpN1qYbrcB56X3rwR+MMjrXAr8YpDH2oBzC7bnAV3p53rwMy445/eM9DM9gu/Lb/XFT/Kj+yzwJqBhgL/XXxdsD/TdeFfB9pnAU/1e4yrgK5U+J5ILjo0Fjx2VPveYgf62wBuB/wTOAiaV8v++Vv+5jWHs+DrJVdjNaRH9G8BHIuIFSReTXPl+SdJPgD+LiCeGeb3FQAOwXVLfvknA1iGes6Pgfh7YFRE9BduQ/BDuSYvcHyW5eppE8p/ql8OfJsf2i2GgeJ4puL8/fQ6SjgI+R1KamJE+3iypriDO/s/ta+dZCPxqkJi+AbRJmkZSWrkvIrYPcQ4HY46IfZKeKzivG0lKHHemt58f5DWGimcxcKuk3oJ9PSQ/dOUq9TM9liG+L5KOJim5vo7kx3cSSUmGiNiYVpddDZwk6Q7ggxHxdIkxFn4HFgPHStpTsK+O5Oq9oueUOvh9iYj96XHTGEBE/FjSF0hKb4sk3Qp8KCI6hj3DGuOqpDEiIroi4mMRcSJJddFbgCvSx+6IiPNIrhyfAP4xfdoLJD/IfY4puL+V5GppdkS0pv+mRwV6CEmaAnwP+DQwNyJagR+QVDcMZzuwoGB7YRlv/WfACcCZETGdpFhPie+7FXjpQA9ExDbgp8B/By5n6GokKIg5TSYzgb4fwW8AF0k6haTK5fvlxpM+dn7B59YaEY1pnFkZ7vtyLcnV9Mnp3/4yCv7uEfGtiHgtyQ97AJ9MHxrqO3rw6f3i2Nzv3Jsj4oIMzmk4h009HRHXRcQrgZNILor+/AjiGvOcGMYISW+Q9PK0YayDpOqgR9JcSavStoYDJMXmvqujh4GzJS2S1EJS5AYgveL9IfAZSdPTBs2XSnp9BcKdDEwBdgLdaenhzUM/5aBbgKuUNCTPJ6luKVUzyRXhnrQx9qNlPPebwJskvT1t4Jwl6dSCx78G/AXwcuDWYV7rAiUN2ZNJ2hp+FhFbASKiHXiQJLl8LyLyg7zG7cAxkv5E0hRJzZLOTB/7B+BvdKhRe46ki8o417KV8H1pJvnu7Uk/t4M/iJJOkPTG9IKhk+QzKvyOXiBppqRjgAEb4gv8P6AjbZBuklQnabmk0zM4p+HsIGkLA0BJR4UzJTWQJLxODp3nuOLEMHYcA3yXJCm0AfeQXH1OIrlSfhp4jqRB8n0AEXEn8G3gUeAhkh+bQleQ/Ig/TlLs/y5JqWNEImIvSQPpLenr/g6wtsSnXwO0A5tJGl+/S5LwSvF/gCZgF/AAScNtqTE/BVxA8rd8juQH65SCQ24lrcKJiBeGeblvkSSl54BXkjRGF7qRJMEMWvJI/4bnkdSRPwP8F/CG9OHPk/w9fyhpL8m5njnQ61TYUN+XjwGnATngX4F/LnjeFOATJJ/LMyQN/3+ZPvZ1kobkLSQ/0t8eKoC0SuhCkvajzelr/hNJx4xKn9NwPg+8Le2xdB1JR4t/TF/n1yQNz58+wrjGNKWNKmZVIem9wCURUYmSzEhj+RXJwMMRjTWRdDZJUl8SEb3DHW821rjEYKNK0jxJr0mL9SeQXMEPV3UzGnG9laRO+ccjfJ0G4APAPzkpWK1yryQbbZNJ+pkvJemGezPw99UMSMnUBycCl4/kx1zSMmA9SdXJ71UmOrPR56okMzMr4qokMzMr4sRgZmZFarKNYfbs2bFkyZJqh2FmVlMeeuihXRExZ7jjajIxLFmyhPXr11c7DDOzmiLp16Uc56okMzMr4sRgZmZFnBjMzKxITbYxDKSrq4v29nY6OzurHUpJGhsbWbBgAQ0NDdUOxcysyLhJDO3t7TQ3N7NkyRIK5l4fkyKC3bt3097eztKlS6sdjplZkUyrkiR9OV0fdcMgj0vSdZI2puunlrI05IA6OzuZNWvWmE8KAJKYNWtWzZRuzGxiybrE8FWSNXm/Nsjj5wPHpf/OBP4vI5heuBaSQp9aitXMquyZDdB2G+S2QstCWHYhHLM8s7fLtMQQEfeSzFk/mIuAr0XiAaBV0ojXC6iWuro6Tj31VE455RROO+007r///mqHZGa17pkNcP/1kN8D0+cnt/dfn+zPSLXbGOZTvP5qe7pvqPV2x6ympiYefvhhAO644w6uuuoq7rnnnipHZWaZGK2r+LbboLEVmlqT7b7bttsyKzVUOzEMVJ8y4HSvklYDqwEWLVo04jdu255j3YYdbNuTZ35rEyuXz2XZvCNdJOpwHR0dzJgxY/gDzaz29F3FN7YWX8W/+v2V/7HObU3eo1Dj9GR/RqqdGNopXgx+AYcWVS8SEWuANQArVqwY0VzhbdtzrLl3My1NDcxraSSX72LNvZtZffbSESWHfD7PqaeeSmdnJ9u3b+fHPx7Rmi9mNlaN5lV8y8Ik8fS9B0BnR7I/I9Ue4LYWuCLtnXQWkEsX8M7Uug07aGlqoKWpgUnSwfvrNuwY0ev2VSU98cQTrFu3jiuuuAKvd2E2DuW2JlfthbK6il92IXTuSZJD9Ca3nXuS/RnJurvqTcBPgRMktUt6t6Q/lPSH6SE/ADYBG0kW2X5flvH02bYnT3NjcWGpubGebXvyFXuPV73qVezatYudO3dW7DXNbIxoWZhctRfK6ir+mOVJFVVTK3RsS26zqLIqkGlVUkRcOszjAfxRljEMZH5rE7l8Fy1Nh0Yd7+3sZn5rU8Xe44knnqCnp4dZs2ZV7DXNbIxYdmHSpgBJSaGzI7mKP+3ybN7vmOWZJoL+qt3GUBUrl89lzb2bgaSksLezm1y+i4tPXzCi1+1rY4BkdPONN95IXV3diOM1szGm7yq+sFfSaZeP6o93liZkYlg2r4XVZy8t6pV08ekLRtwrqaenp0IRmtmYN8pX8aNpQiYGSJJDJbunmpmNF9XulWRmZmOME4OZmRVxYjAzsyJODGZmVsSJwczMijgxVJAkLr/80ACX7u5u5syZw1ve8pYqRmU2gT2zAe66Fr7/vuQ2w6mqxxMnhgqaOnUqGzZsIJ9Ppta48847mT9//jDPMrNMVGEdg/Fi4iaGjK4kzj//fP71X/8VgJtuuolLLx1yVhAzy0rhDKialNw2tib7bUgTMzFkeCVxySWXcPPNN9PZ2cmjjz7KmWce8UqlZjYSozkD6jgzMRNDhlcSJ598Mlu2bOGmm27iggsuqECwZnZERnMG1HFmYiaGjK8kVq1axYc+9CFXI5lVUxXWMRgvJuZcSRmviPSud72LlpYWXv7yl3P33XdX5DXNrEzjfAbULE3MxJDxXOoLFizgAx/4QEVey8xGYBzPgJqliVmVlNGKSPv27Tts3znnnMPtt98+otc1MxtNE7PEAL6SMDMbxMQsMZiZ2aCcGMzMrMi4SgwRUe0QSlZLsZrZxDJuEkNjYyO7d++uiR/ciGD37t00NjZWOxQzs8OMm8bnBQsW0N7ezs6dO6sdSkkaGxtZsGBBtcMwMzvMuEkMDQ0NLF26tNphmJnVvHFTlWRmZpXhxGBmZkWcGMzMrIgTg5mZFXFiMDOzIk4MZmZWJPPEIGmlpCclbZT04QEeXyTpLkm/kPSoJC97ZmZWRZkmBkl1wA3A+cCJwKWSTux32P8CbomIVwCXAH+fZUxmZja0rEsMZwAbI2JTRLwI3Axc1O+YAPrW2WwBns44JjMzG0LWI5/nA4ULKbcDZ/Y75mrgh5LeD0wF3pRxTGY2Wp7ZULy05rILvQ5KDci6xKAB9vWf5e5S4KsRsQC4APi6pMPikrRa0npJ62tlPiSzCe2ZDckSuvk9MH1+cnv/9cl+G9OyTgztwMKC7QUcXlX0buAWgIj4KdAIzO7/QhGxJiJWRMSKOXPmZBSumVVM223Q2JosnatJyW1ja7LfxrSsE8ODwHGSlkqaTNK4vLbfMU8B5wJIWkaSGFwkMKt1ua3QOL14X+P0ZL+NaZkmhojoBq4E7gDaSHofPSbpGkmr0sP+DPh9SY8ANwHvjFpYVMHMhtayEDo7ivd1diT7bUzLfNrtiPgB8IN++/53wf3HgddkHYeZjbJlFyZtCpCUFDo7oHMPnHZ5deOyYXnks5ll45jl8Or3J20LHduS21e/372SasC4WajHzMagY5Y7EdQgJwazWuOxAZYxJwazWtI3NqCxtXhswHBVNE4mVgYnBrNSVevHtfB9n98C0+Yl9fVw6LbttsFjOdJkYhOWG5/NSlGtUbz93/eFXbDzcdj37KFjhhsb4IFmViYnBrNSVOvHtf/7Tp0NTIJdTx46ZrixAR5oZmVyYjArRbV+XPu/7+wTIHqTkkP0JiWJzj1JtdZgPNDMyuTEYFaKav249n/faUfD3JOSkkOpYwOWXZgkj/ye0pOJTWhODGalqNaP60DvO6kOzv8U/NbfwxuuGr4B2QPNrEyqxWmJVqxYEevXr692GDbRjIVeSe5qaiMg6aGIWDHcce6ualaqao3i9ehhG2WuSjIzsyJODGZmVsSJwczMipScGCT9din7zMystpVTYriqxH1mZlbDhu2VJOl84AJgvqTrCh6aDnRnFZiZmVVHKd1VnwbWA6uAhwr27wX+NIugzMyseoZNDBHxCPCIpG9FRBeApBnAwoh4PusAzcxsdJXTxnCnpOmSZgKPAF+R9NmM4jIzsyopZ+RzS0R0SHoP8JWI+KikR7MKzKwmefoKGwfKKTHUS5oHvB24PaN4zGpXtRbzMauwchLDx4A7gI0R8aCklwD/lU1YZjXIK6XZOFFSVZKkOpLG5pP79kXEJuCtWQVmVnNyW5OSQiGvlGY1qKQSQ0T0kHRXNbPBeKU0GyfKaXy+X9IXgG8DL/TtjIifVzwqs7Gg3IbkZRcmbQqQlBQ6O5JFdk67fHTiNauQkhfqkXTXALsjIt5Y2ZCG54V6LHN9DcmNrcU/8sOtfOZeSTaGVXyhnoh4w8hCMhtDhvsBL2xIhkO3bbcN/UPvRXVsHChndtUWSZ+VtD799xlJLVkGZ5aJUrqV5rYmJYVCbki2CaKc7qpfJpkf6e3pvw7gK8M9SdJKSU9K2ijpw4Mc83ZJj0t6TNK3yojJrHyldCt1Q7JNYOU0Pr80Igq7p35M0sNDPSHt5noDcB7QDjwoaW1EPF5wzHEk03e/JiKel3R0GTGZla+UbqVuSLYJrJwSQ17Sa/s2JL0GyA/znDNIBsRtiogXgZuBi/od8/vADX0T8kXEs2XEZFa+UkoDxyxPGpqbWqFjW3I7XMOz2ThRTonhvcCNabuCgOeAdwzznPlAYaVsO3Bmv2OOB5D0E6AOuDoi1pURl1l5Si0NuCHZJqhyeiU9DJwiaXq63THMUyBJIIe91AAxHAecAywA7pO0PCL2FL2QtBpYDbBo0aJSwzY7XF9poLBX0mmXOwmYpUpODJJ+BTwA3AfcCzw+9DOApIRQ2Fq3gGThn/7HPJCu9bBZ0pMkieLBwoMiYg2wBpJxDKXGbTYglwbMBlVOG8OJwBeBWcCnJW2SdOswz3kQOE7SUkmTgUuAtf2O+T7wBgBJs0mqljaVEZeZmVVQOYmhB+hKb3uBHcCQDcUR0Q1cSTIraxtwS0Q8JukaSX1zL90B7Jb0OHAX8OcRsbu80zAzs0opZ0qM/cAvgc8C/17NH29PiWFmVr5Sp8Qop8RwKUnbwvuAmyV9TNK5RxqgmZmNTeX0SvoX4F8k/QZwPvAnwF8ATRnFZmZmVVDOXEnfS3smfR6YClwBzMgqMDMzq45yBrh9Avh5umjPYSSdFxF3ViYsmxA8RbXZmFRyiSEiHhwsKaQ+WYF4bKIoZYZTM6uKckoMwxlolLPZwI50vYOhuARiVhHl9EoajkcjW+kqvd6BSyBmFVPJEoNZopQr95aFyY93X0kBRrbeQRYlELMJqqQSg6RJkl49zGFbRh6O1bxSr9yXXZjMaJrfA9Gb3HbuSfYfCa+4ZlYxJSWGiOgFPjPMMf+jIhFZbStldTSo/HoHXnHNrGLKqUr6oaS3Av8cpc6jYRNPKauj9ankDKdecc2sYspJDB8kGdjWIylP0gspImL60E+zCaVuCmz8EfS8mPxAzz4B6iZnf+XuNRbMKqacKTGaswzExoFnNiTVQgc6YEozvJiHLffBzJfCaX+V/ft7jQWziihnSgxJukzSX6XbCyWdkV1oVnPaboMZS2HJ66ChCXq7YMp0mH6sf7DNakg5VUl/T7IOwxuBjwP7gBuA0zOIy2pRX/uCJsG0o5N90ZuUIsysZpQzwO3MiPgjoBMgIp4HJmcSldUm9wwyGxfKSQxdkupIRzhLmkNSgjBLVHpsgplVRTmJ4TrgVuBoSX8D/AdwbSZRWW2q9NgEM6uKcnolfVPSQ8C5JF1Vfysi2jKLzGqTewaZ1bySE4Okr0fE5cATA+wzM7NxopyqpJMKN9L2hldWNhwzM6u2YUsMkq4C/hJoktTBoXUXXgTWZBib1RKvhWA2bgybGCLiWuBaSddGxFWjEJPVmE0bHqDzns+zu6eJ+qNaOb57B7Puv94Nz2Y1qpyqpI945LP117Y9x3/efRMdMZW6o2ZwoDt4Yutu9m9vg1v/EO661ovlmNWYchLDDcCrgN9Jt/tGPtsEtm7DDub2Pks0TkcSM2IPLz3wGLn9XUB4JTWzGuSRzzYi2/bkyR91LFN69gEwM7+F3rpGXuzphcaWwddjMLMxq5y5kjzyeTyocCPx/NYmHu45m3N23wzAlK4OXqSepkndyZTb4JXUzGrMSEc+/20mUVk2Sl12swwrl8/lv7SYu2ddQmddM92I7oD6xWcemkjP8yWZ1RSPfJ5ICpfdhEO3bbcdcalh2bwWVp+9lHUbjuKf65Zw8qw3smr/92htbk7mS/JKamY1p5yqJIAdwH3p85oknRYRP698WJaJwZbd3P5o0nvoCKuXls1rYdm8lnTrFHhmsVdSM6th5UyJ8XHgncCvSNsZ0ts3DvO8lcDngTrgnyLiE4Mc9zbgO8DpEbG+1LisDC0Lk+qjvpICwO5NkHsKWhcVVy8NMQahbXuOdRt2sG1PnvmtTaxcPrcgMeD5ksxqXDltDG8HXhoR50TEG9J/wyWFOpIurecDJwKXSjpxgOOagT8GflZGPFaugabF3tkGc5YlyUKThu1F1LY9x5p7N5PLdzGvpZFcvos1926mbXtulE/GzLJSTmLYALQOe1SxM4CNEbEpIl4EbgYuGuC4jwOfIu0KaxkZaFrsGYth5tLi44boRbRuww5amhpoaWpgknTw/roNO0bhBMxsNJTTxnAt8AtJG4ADfTsjYtUQz5kPFP7CtANnFh4g6RXAwoi4XdKHBnshSauB1QCLFi0qI2wr0r+a565rD69eGqIX0bY9eea1NLJrXycbn32Bjs4umqfU03JUQ8aBm9loKScx3Ah8EvglpY9f0AD74uCD0iTgcyRtF0OKiDWkk/atWLEihjncSrXswqRNAZKSwjC9iOa3NrFl1z6e3LGPKfWTaJ5ST0dnNx35btq254rbGsysJpVTlbQrIq6LiLsi4p6+f8M8px0ovPRcADxdsN0MLAfulrQFOAtYK2lFGXHZSJS56trK5XN5ckcyynlK/SQOdCfXCMfPnebqJLNxopwSw0OSrgXWUlyVNFR31QeB4yQtBbYBl3BoriUiIgfM7tuWdDfwIfdKGmVl9CJaNq+FBTOa6Mh3se9AD9Ma6znp2OnMmjaFbXvyGQdqZqOhnMTwivT2rIJ9Q3ZXjYhuSVcCd5B0V/1yRDwm6RpgfUSsLTdgq7xhu5/2c9KxLeTyXbQ0HWpXyOW7mN/aNBrhmlnGFFF71fUrVqyI9etdqBiJvmTw+PYcW3fnOX7uNBbPnsrezm5y+S5Wn7100OTQ12W1pamB5sb6kp5jZtUn6aGIGLaqvpQV3C6LiG9I+uBAj0fEZ48kQKuewh/23P4uEPzns/uY1ljPnOZGIOmWOtiP/KFpMA6VMi4+fYGTgtk4UUpV0tT0tjnLQKxEFZgdtXAswt4D3UxvrOdAdy8bd77AnOZGmhvrh20vKJ4Gw8zGk1KW9vxiOoK5IyI+Nwox2WD6ZkdtbB3R9BWPPZ2jJfckL9v/H/xR1zM8VzeXR6a9jrbOxQDs7ex2e4HZBFZS43NE9EhaRTLmwKqljNlR27bn+MZPf819G3cz46gGls1rJpfv4lPrnkQ7HuP8nn/hQP10nq+bw6QXc5yX+w60/Da5/Exy+S4uPn3BKJ+cmY0V5fRKul/SF4BvAy/07fTsqqNosNlR+01f0deGsGnnPlqbko/44a05ls46isef7uAdXf/BszTRrSaaGup4IabzXJc4u/cBNjWd4fYCswmunMTw6vT2moJ9w86uahU00OyoA0xf0deG0NHZxYGuXl7s6aW3N9ieyzMJMZ9d7J18NPmuHnojmDq5nilHzWDVkl447/hRPikzG2vKWajnDVkGYiUocfqKbXvyNNRBR74bSEYo7znwIt29wbQp9TyrORxbf4C6uqnU14k506YwXS9Ay9L+72hmE1DJU2JImivpS5L+Ld0+UdK7swvNDlPi9BXzW5t4/Om9zJ42GYCe3qAngkkSPRH8dPJraOrZy3T20XmgCx3I8ZJp3UniMbMJr5yqpK8CXwE+km7/J0l7w5cqHJMNZYjpK/p6ID32dI4tu/czt3kyrU31PNPRSVdP0FAXTJvcwLwTXskP2ifz8r33sqh+N8techKtp/+2F9cxM6C8xDA7Im6RdBUcnO6iJ6O4rEyFg9aWzZvOr3e/wPZcJwFMm9LA3OZJ7NrXxQtdvfT0BtMWn8rD+ZM44+ylHO2GZjMrUM7sqi9ImkU6bbakswAv2zVG9F9A5/QlM6mbNInG+joWzzqK5qbJzJw2mbnTpvBIe46WpgZPYWFmAyqnxPBBkplVXyLpJ8Ac4G2ZRGVl61tAp8+c5kZmTG2gI9+Vjm5uYPn86cycOoXtuU7+1L2PzGwQ5SSGx4Fbgf3AXuD7JO0MVgX9RzRPrhN7O7uLZjyd3tjA9MYGzjnh6IP7PAuqmQ2nnKqkrwG/AfwtcD1wHPD1LIKyofW1J+TyXcxraSSX72JHxwF+vfsFcvkueiPI5buYOXUys6dNKdqXy3excvncap+CmY1h5ZQYToiIUwq275L0SKUDsuEVticAye3Mo+jq7qGlqeFgKeIvVp5w8HjPgmpmpSonMfxC0lkR8QCApDOBn2QTlg2lf3sCQHNjPdtz3QO2HTgRmFk5ykkMZwJXSHoq3V4EtEn6JRARcXLFo7MBzW9tOmwFNc+IamaVUk5iWJlZFDaswsbmyXViR8cBmHlU0QpqnhHVzCqhnLmSfp1lIHZI/x5Hx8+dyr+37aSlqYF5LY3s7eymN4Ku7h6257rddmBmFVVOicFGQeEI5r4eR9f/6FeccMy0osbmxbOm0tLU4PEIZlZx5XRXtVHQfwRzS1MD3b3B9lxn0XGlLL9pZnYkXGIYY/r3ONq5t5P9L3azo6MTJF42Zypzmhvd2GxmmXGJYYyZ39rE3s5kHYWdezv5+VN7mFI/iYY60ZHv4ue/3sPmnfs8UM3MMuPEMMasXD734Ajljc/uA2BKQx3LhCNQAAAL70lEQVRnvWQmLU0NdPX28szeA54Az8wy46qkMWbZvBZWn72UdRt2sGPvAeY2T+G4udOYPa2R44DeSNobnBTMLCtODGPQsnktB3/4PZDNzEabq5LGsMJqJU+CZ2ajxYlhDOurVmppamB7rtOL65jZqHBV0hhXWK1kZjYaMi8xSFop6UlJGyV9eIDHPyjpcUmPSvqRpMVZx2RmZoPLtMQgqQ64ATgPaAcelLQ2Ih4vOOwXwIqI2C/pvcCngIuzjGss6JsP6bGnc3R0dtPSVM+J81pYuXyuSwhmVlVZlxjOADZGxKaIeBG4Gbio8ICIuCsi9qebDwDjforQvvmQNu/cR/tzeTryXTy1ez9bdu1jzb2badueq3aIZjaBZZ0Y5gNbC7bb032DeTfwb5lGNAb0zYf0zN4DTGmYREtTA40NdTzTcYCWpgbWbdhR7RDNbALLuvFZA+yLAQ+ULgNWAK8f5PHVwGqARYsWVSq+quibD2lfZzfTptQBMKV+Eh2dXZ4cz8yqLusSQzuwsGB7AfB0/4MkvQn4CLAqIg4M9EIRsSYiVkTEijlz5mQS7Gjpmw9pWmM9B7p7ATjQ3cv0xgYPYDOzqss6MTwIHCdpqaTJwCXA2sIDJL0C+CJJUng243jGhL6Ba8c0T+FAVy+5fBedXT0cM32KB7CZWdVlWpUUEd2SrgTuAOqAL0fEY5KuAdZHxFrg74BpwHckATwVEauyjKvaCudD2t/Vc7BX0pLZ09wrycyqThEDVvmPaStWrIj169dXOwwzs5oi6aGIWDHccR75nJH+6za7JGBmtcJzJWWgb5xCLt91cN1mj08ws1rhxJCBgdZt9vgEM6sVTgwZ2LYnT3NjcS2dxyeYWa1wYshA4brNfTw+wcxqhRNDBrzAjpnVMvdKOkJD9ToqHKfQ9/jFpy9wryQzqwlODEegr9dRS1NDUa+jwtXVvMCOmdUqVyUdAfc6MrPxzInhCLjXkZmNZ04MR8C9jsxsPHNiOALudWRm45kTwxHo63XU0tTA9lwnLU0NRQ3PZma1zL2SjpB7HZnZeOUSg5mZFXFiMDOzIk4MZmZWxInBzMyKODGYmVkRJwYzMyvixGBmZkWcGMzMrIgTg5mZFXFiMDOzIk4MZmZWxInBzMyKODGYmVkRJwYzMyvixGBmZkWcGMzMrEjmiUHSSklPStoo6cMDPD5F0rfTx38maUnWMZmZ2eAyXcFNUh1wA3Ae0A48KGltRDxecNi7gecj4mWSLgE+CVycZVxt23Os27CDbXvyzG9tYuXyuV6NzcwslXWJ4QxgY0RsiogXgZuBi/odcxFwY3r/u8C5kpRVQG3bc6y5dzO5fBfzWhrJ5btYc+9m2rbnsnpLM7OaknVimA9sLdhuT/cNeExEdAM5YFZWAa3bsIOWpgZamhqYJB28v27Djqze0syspmSdGAa68o8jOAZJqyWtl7R+586dRxzQtj15mhuLa9CaG+vZtid/xK9pZjaeZJ0Y2oGFBdsLgKcHO0ZSPdACPNf/hSJiTUSsiIgVc+bMOeKA5rc2sbezu2jf3s5u5rc2HfFrmpmNJ1knhgeB4yQtlTQZuARY2++YtcA70vtvA34cEYeVGCpl5fK55PJd5PJd9EYcvL9y+dys3tLMrKZkmhjSNoMrgTuANuCWiHhM0jWSVqWHfQmYJWkj8EHgsC6tlbRsXgurz15KS1MD23OdtDQ1sPrspe6VZGaWUoYX55lZsWJFrF+/vtphmJnVFEkPRcSK4Y7zyGczMyvixGBmZkWcGMzMrIgTg5mZFXFiMDOzIk4MZmZWxInBzMyK1OQ4Bkk7gV9X4KVmA7sq8Dq1wuc7vk2084WJd84jPd/FETHsnEI1mRgqRdL6UgZ7jBc+3/Ftop0vTLxzHq3zdVWSmZkVcWIwM7MiEz0xrKl2AKPM5zu+TbTzhYl3zqNyvhO6jcHMzA430UsMZmbWjxODmZkVmRCJQdJKSU9K2ijpsIWAJE2R9O308Z9JWjL6UVZOCef7QUmPS3pU0o8kLa5GnJUy3PkWHPc2SSGpprs3lnK+kt6efsaPSfrWaMdYSSV8nxdJukvSL9Lv9AXViLNSJH1Z0rOSNgzyuCRdl/49HpV0WsWDiIhx/Q+oA34FvASYDDwCnNjvmPcB/5DevwT4drXjzvh83wAcld5/73g/3/S4ZuBe4AFgRbXjzvjzPQ74BTAj3T662nFnfL5rgPem908EtlQ77hGe89nAacCGQR6/APg3QMBZwM8qHcNEKDGcAWyMiE0R8SJwM3BRv2MuAm5M738XOFeSRjHGShr2fCPirojYn24+ACwY5RgrqZTPF+DjwKeAztEMLgOlnO/vAzdExPMAEfHsKMdYSaWcbwDT0/stwNOjGF/FRcS9wHNDHHIR8LVIPAC0SppXyRgmQmKYD2wt2G5P9w14TCTrVOeAWaMSXeWVcr6F3k1y9VGrhj1fSa8AFkbE7aMZWEZK+XyPB46X9BNJD0haOWrRVV4p53s1cJmkduAHwPtHJ7SqKff/eNnqK/liY9RAV/79++iWckytKPlcJF0GrABen2lE2RryfCVNAj4HvHO0AspYKZ9vPUl10jkkpcH7JC2PiD0Zx5aFUs73UuCrEfEZSa8Cvp6eb2/24VVF5r9XE6HE0A4sLNhewOFFzYPHSKonKY4OVZQby0o5XyS9CfgIsCoiDoxSbFkY7nybgeXA3ZK2kNTJrq3hBuhSv8//EhFdEbEZeJIkUdSiUs733cAtABHxU6CRZLK58aqk/+MjMRESw4PAcZKWSppM0ri8tt8xa4F3pPffBvw40laeGjTs+aZVK18kSQq1XP8Mw5xvROQiYnZELImIJSRtKqsiYn11wh2xUr7P3yfpYICk2SRVS5tGNcrKKeV8nwLOBZC0jCQx7BzVKEfXWuCKtHfSWUAuIrZX8g3GfVVSRHRLuhK4g6SHw5cj4jFJ1wDrI2It8CWS4udGkpLCJdWLeGRKPN+/A6YB30nb2J+KiFVVC3oESjzfcaPE870DeLOkx4Ee4M8jYnf1oj5yJZ7vnwH/KOlPSapU3lnDF3ZIuomkGnB22m7yUaABICL+gaQd5QJgI7Af+L2Kx1DDfz8zM8vARKhKMjOzMjgxmJlZEScGMzMr4sRgZmZFnBjMaoikVknvq3YcNr45MZilJNVVO4Y+6UDLgbSSTPpolhknBpswJH1f0kPpVNSr0337JF0j6WfAqyS9UtI96XF39E1OJun3JT0o6RFJ35N01CDvUSdpUzr4qFVSr6Sz08fuk/QySTPTWB5N5zI6OX38aklrJP0Q+JqkkyT9P0kPp8ceB3wCeGm67+9G4+9mE4/HMdiEIWlmRDwnqYlkRO3rgV3AxRFxi6QG4B7goojYKeli4Dcj4l2SZvUNEpP018COiLh+kPdZRzLoainJ4KTvA58GnoiIpZKuB3ZFxMckvRH4bEScKulq4ELgtRGRT497ICK+mY76rQPmArdHxPKM/kxm43/ks1mBP5b039P7C0nmD+oBvpfuO4FkXqU70xHhdUDfVAPL04TQSjJq/I4h3uc+kjn1lwLXkkyDfQ9JMgJ4LfBWgIj4saRZklrSx9ZGRD69/1PgI5IWAP8cEf9Vu7PBWy1xVZJNCJLOAd4EvCoiTiFZyKYR6IyInr7DgMci4tT038sj4s3pY18FroyIlwMfS587mPuA15GsJfADkmRyDslCQX3v019f0f2FgzsivgWsAvLAHWnpwixzTgw2UbQAz0fEfkm/QTLLan9PAnPSqZuR1CDppPSxZmB7Wt30u8O818+AVwO9EdEJPAz8AUnCgCRB/G76HueQVCt19H8RSS8BNkXEdSQTp50M7E1jMcuME4NNFOuAekmPkqzm9kD/A9IVwt4GfFLSIyQ/6K9OH/4rkh/8O4EnhnqjdBrzrQXvcR/Jj/kv0+2rgRVpLJ/g0My+/V0MbJD0MPAbJKt27QZ+ImmDG58tK258NjOzIi4xmJlZEfdKMjtCkj4C/Ha/3d+JiL+pRjxmleKqJDMzK+KqJDMzK+LEYGZmRZwYzMysiBODmZkVcWIwM7MiTgxmZlbk/wNL+qSjDplRaAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(features_train_rus[label_train_rus == \"B\"].area_worst, \n",
    "            features_train_rus[label_train_rus == \"B\"].perimeter_worst, \n",
    "            label=\"B\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.scatter(features_train_rus[label_train_rus == \"M\"].area_worst, \n",
    "            features_train_rus[label_train_rus == \"M\"].perimeter_worst, \n",
    "            label=\"M\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.title(\"Tissue malignancy by cell measurements\")\n",
    "plt.xlabel(\"area_worst\")\n",
    "plt.ylabel(\"perimeter_worst\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "El gráfico nos comunica que sí hay un balance en las clases del conjunto de datos, pero hay significativamente menos datos para entrenar."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "                   intercept_scaling=1, l1_ratio=None, max_iter=100,\n",
       "                   multi_class='auto', n_jobs=None, penalty='l2',\n",
       "                   random_state=11, solver='lbfgs', tol=0.0001, verbose=0,\n",
       "                   warm_start=False)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr = sklearn.linear_model.LogisticRegression(random_state=11)\n",
    "lr.fit(features_train_rus, label_train_rus)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr_predictions = lr.predict(features_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           B       0.99      0.97      0.98       108\n",
      "           M       0.73      0.89      0.80         9\n",
      "\n",
      "    accuracy                           0.97       117\n",
      "   macro avg       0.86      0.93      0.89       117\n",
      "weighted avg       0.97      0.97      0.97       117\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(sklearn.metrics.classification_report(label_test, lr_predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Logramos aumentar el recall de la clase M a a costa de disminuir la precisión, pero nuestras métricas en general aumentaron."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sobremuestreo de la clase minoritaria"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Otro enfoque simple para resolver el desbalance de las clases es remuestrar la clase minoritaria al multiplicar observaciones de esa clase de manera aleatoria hasta alcanzar la frecuencia de la clase mayoritaria.\n",
    "\n",
    "Esta técnica está implementada en la clase imblearn.over_sampling.RandomOverSampler()."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "ros = imblearn.over_sampling.RandomOverSampler(random_state=11) # Instanciamos la clase para sobremuestrear nuestro conjunto de datos.\n",
    "features_train_ros, label_train_ros = ros.fit_resample(features_train, label_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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.scatter(features_train_ros[label_train_ros == \"B\"].area_worst, \n",
    "            features_train_ros[label_train_ros == \"B\"].perimeter_worst, \n",
    "            label=\"B\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.scatter(features_train_ros[label_train_ros == \"M\"].area_worst, \n",
    "            features_train_ros[label_train_ros == \"M\"].perimeter_worst, \n",
    "            label=\"M\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.title(\"Tissue malignancy by cell measurements\")\n",
    "plt.xlabel(\"area_worst\")\n",
    "plt.ylabel(\"perimeter_worst\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Se evidencia que tenemos una cantidad significativa de datos, pero la clase minoritaria fue aumentada sólo a costa de exactamente los mismos datos de la clase."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "                   intercept_scaling=1, l1_ratio=None, max_iter=100,\n",
       "                   multi_class='auto', n_jobs=None, penalty='l2',\n",
       "                   random_state=11, solver='lbfgs', tol=0.0001, verbose=0,\n",
       "                   warm_start=False)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr = sklearn.linear_model.LogisticRegression(random_state=11)\n",
    "lr.fit(features_train_ros, label_train_ros)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr_predictions = lr.predict(features_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           B       1.00      0.96      0.98       108\n",
      "           M       0.69      1.00      0.82         9\n",
      "\n",
      "    accuracy                           0.97       117\n",
      "   macro avg       0.85      0.98      0.90       117\n",
      "weighted avg       0.98      0.97      0.97       117\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(sklearn.metrics.classification_report(label_test, lr_predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Aumentamos el recall de la clase M y aumentamos las métricas generales del experimento. Este método al parecer es mejor que el anterior, porque en ese se pierde mucha información al eliminar datos."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## _Synthetic Minority Oversampling Technique_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "El sobremuestreo al azar no agrega información nueva al conjunto de datos. La técnica que exploraremos ahora sintetiza nuevos datos desde las observaciones con las que cuenta el conjunto de datos.\n",
    "\n",
    "SMOTE crea nuevas observaciones al trazar lineas entre observaciones de la clase minoritaria y agrega al azar una observación sobre esa línea.\n",
    "\n",
    "Esté método está implementado en la clase imblearn.over_sampling.SMOTE()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "sm = imblearn.over_sampling.SMOTE(random_state=11)\n",
    "features_train_smote, label_train_smote = sm.fit_resample(features_train, label_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "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.scatter(features_train_smote[label_train_smote == \"B\"].area_worst, \n",
    "            features_train_smote[label_train_smote == \"B\"].perimeter_worst, \n",
    "            label=\"B\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.scatter(features_train_smote[label_train_smote == \"M\"].area_worst, \n",
    "            features_train_smote[label_train_smote == \"M\"].perimeter_worst, \n",
    "            label=\"M\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tenemos un conjunto de datos balanceados y las observaciones sintetizadas siguen la silueta del espacio original."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "scrolled": true
   },
   "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>radius_mean</th>\n",
       "      <th>texture_mean</th>\n",
       "      <th>perimeter_mean</th>\n",
       "      <th>area_mean</th>\n",
       "      <th>smoothness_mean</th>\n",
       "      <th>compactness_mean</th>\n",
       "      <th>concavity_mean</th>\n",
       "      <th>concave points_mean</th>\n",
       "      <th>symmetry_mean</th>\n",
       "      <th>fractal_dimension_mean</th>\n",
       "      <th>...</th>\n",
       "      <th>radius_worst</th>\n",
       "      <th>texture_worst</th>\n",
       "      <th>perimeter_worst</th>\n",
       "      <th>area_worst</th>\n",
       "      <th>smoothness_worst</th>\n",
       "      <th>compactness_worst</th>\n",
       "      <th>concavity_worst</th>\n",
       "      <th>concave points_worst</th>\n",
       "      <th>symmetry_worst</th>\n",
       "      <th>fractal_dimension_worst</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "      <td>249.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.667711</td>\n",
       "      <td>0.435546</td>\n",
       "      <td>0.696743</td>\n",
       "      <td>0.578279</td>\n",
       "      <td>0.147886</td>\n",
       "      <td>0.009209</td>\n",
       "      <td>0.055266</td>\n",
       "      <td>0.045299</td>\n",
       "      <td>0.082338</td>\n",
       "      <td>0.032053</td>\n",
       "      <td>...</td>\n",
       "      <td>0.581314</td>\n",
       "      <td>0.541189</td>\n",
       "      <td>0.624648</td>\n",
       "      <td>0.468427</td>\n",
       "      <td>0.007446</td>\n",
       "      <td>0.033393</td>\n",
       "      <td>0.029157</td>\n",
       "      <td>0.058714</td>\n",
       "      <td>0.040365</td>\n",
       "      <td>0.009861</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.215136</td>\n",
       "      <td>0.132995</td>\n",
       "      <td>0.212213</td>\n",
       "      <td>0.252425</td>\n",
       "      <td>0.344231</td>\n",
       "      <td>0.025292</td>\n",
       "      <td>0.179230</td>\n",
       "      <td>0.185791</td>\n",
       "      <td>0.220165</td>\n",
       "      <td>0.131703</td>\n",
       "      <td>...</td>\n",
       "      <td>0.195194</td>\n",
       "      <td>0.195923</td>\n",
       "      <td>0.193755</td>\n",
       "      <td>0.233539</td>\n",
       "      <td>0.028406</td>\n",
       "      <td>0.086725</td>\n",
       "      <td>0.116995</td>\n",
       "      <td>0.092291</td>\n",
       "      <td>0.151395</td>\n",
       "      <td>0.028081</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.313069</td>\n",
       "      <td>0.188797</td>\n",
       "      <td>0.361739</td>\n",
       "      <td>0.192463</td>\n",
       "      <td>0.000896</td>\n",
       "      <td>0.000983</td>\n",
       "      <td>0.001127</td>\n",
       "      <td>0.001137</td>\n",
       "      <td>0.004615</td>\n",
       "      <td>0.000104</td>\n",
       "      <td>...</td>\n",
       "      <td>0.328625</td>\n",
       "      <td>0.156250</td>\n",
       "      <td>0.343339</td>\n",
       "      <td>0.157420</td>\n",
       "      <td>0.000910</td>\n",
       "      <td>0.000171</td>\n",
       "      <td>0.000425</td>\n",
       "      <td>0.001129</td>\n",
       "      <td>0.006040</td>\n",
       "      <td>0.000906</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.475986</td>\n",
       "      <td>0.339419</td>\n",
       "      <td>0.533241</td>\n",
       "      <td>0.344942</td>\n",
       "      <td>0.001014</td>\n",
       "      <td>0.001145</td>\n",
       "      <td>0.001567</td>\n",
       "      <td>0.005651</td>\n",
       "      <td>0.005312</td>\n",
       "      <td>0.000624</td>\n",
       "      <td>...</td>\n",
       "      <td>0.419492</td>\n",
       "      <td>0.453965</td>\n",
       "      <td>0.447634</td>\n",
       "      <td>0.289530</td>\n",
       "      <td>0.001212</td>\n",
       "      <td>0.002153</td>\n",
       "      <td>0.002642</td>\n",
       "      <td>0.001624</td>\n",
       "      <td>0.007367</td>\n",
       "      <td>0.001286</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.625503</td>\n",
       "      <td>0.461411</td>\n",
       "      <td>0.641366</td>\n",
       "      <td>0.507100</td>\n",
       "      <td>0.001319</td>\n",
       "      <td>0.001607</td>\n",
       "      <td>0.002187</td>\n",
       "      <td>0.008431</td>\n",
       "      <td>0.005974</td>\n",
       "      <td>0.001080</td>\n",
       "      <td>...</td>\n",
       "      <td>0.568738</td>\n",
       "      <td>0.540995</td>\n",
       "      <td>0.614071</td>\n",
       "      <td>0.428760</td>\n",
       "      <td>0.001318</td>\n",
       "      <td>0.004059</td>\n",
       "      <td>0.003863</td>\n",
       "      <td>0.002114</td>\n",
       "      <td>0.009598</td>\n",
       "      <td>0.007429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.859652</td>\n",
       "      <td>0.535270</td>\n",
       "      <td>0.922920</td>\n",
       "      <td>0.777073</td>\n",
       "      <td>0.009767</td>\n",
       "      <td>0.002439</td>\n",
       "      <td>0.010149</td>\n",
       "      <td>0.009413</td>\n",
       "      <td>0.006981</td>\n",
       "      <td>0.002928</td>\n",
       "      <td>...</td>\n",
       "      <td>0.703861</td>\n",
       "      <td>0.633737</td>\n",
       "      <td>0.741987</td>\n",
       "      <td>0.584739</td>\n",
       "      <td>0.001529</td>\n",
       "      <td>0.006248</td>\n",
       "      <td>0.005880</td>\n",
       "      <td>0.182644</td>\n",
       "      <td>0.011780</td>\n",
       "      <td>0.008121</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.626141</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.976928</td>\n",
       "      <td>0.113723</td>\n",
       "      <td>0.893114</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.772892</td>\n",
       "      <td>0.703186</td>\n",
       "      <td>...</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.963038</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.139306</td>\n",
       "      <td>0.336108</td>\n",
       "      <td>0.572187</td>\n",
       "      <td>0.245207</td>\n",
       "      <td>0.778905</td>\n",
       "      <td>0.173421</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       radius_mean  texture_mean  perimeter_mean   area_mean  smoothness_mean  \\\n",
       "count   249.000000    249.000000      249.000000  249.000000       249.000000   \n",
       "mean      0.667711      0.435546        0.696743    0.578279         0.147886   \n",
       "std       0.215136      0.132995        0.212213    0.252425         0.344231   \n",
       "min       0.313069      0.188797        0.361739    0.192463         0.000896   \n",
       "25%       0.475986      0.339419        0.533241    0.344942         0.001014   \n",
       "50%       0.625503      0.461411        0.641366    0.507100         0.001319   \n",
       "75%       0.859652      0.535270        0.922920    0.777073         0.009767   \n",
       "max       1.000000      0.626141        1.000000    1.000000         0.976928   \n",
       "\n",
       "       compactness_mean  concavity_mean  concave points_mean  symmetry_mean  \\\n",
       "count        249.000000      249.000000           249.000000     249.000000   \n",
       "mean           0.009209        0.055266             0.045299       0.082338   \n",
       "std            0.025292        0.179230             0.185791       0.220165   \n",
       "min            0.000983        0.001127             0.001137       0.004615   \n",
       "25%            0.001145        0.001567             0.005651       0.005312   \n",
       "50%            0.001607        0.002187             0.008431       0.005974   \n",
       "75%            0.002439        0.010149             0.009413       0.006981   \n",
       "max            0.113723        0.893114             1.000000       0.772892   \n",
       "\n",
       "       fractal_dimension_mean           ...             radius_worst  \\\n",
       "count              249.000000           ...               249.000000   \n",
       "mean                 0.032053           ...                 0.581314   \n",
       "std                  0.131703           ...                 0.195194   \n",
       "min                  0.000104           ...                 0.328625   \n",
       "25%                  0.000624           ...                 0.419492   \n",
       "50%                  0.001080           ...                 0.568738   \n",
       "75%                  0.002928           ...                 0.703861   \n",
       "max                  0.703186           ...                 1.000000   \n",
       "\n",
       "       texture_worst  perimeter_worst  area_worst  smoothness_worst  \\\n",
       "count     249.000000       249.000000  249.000000        249.000000   \n",
       "mean        0.541189         0.624648    0.468427          0.007446   \n",
       "std         0.195923         0.193755    0.233539          0.028406   \n",
       "min         0.156250         0.343339    0.157420          0.000910   \n",
       "25%         0.453965         0.447634    0.289530          0.001212   \n",
       "50%         0.540995         0.614071    0.428760          0.001318   \n",
       "75%         0.633737         0.741987    0.584739          0.001529   \n",
       "max         0.963038         1.000000    1.000000          0.139306   \n",
       "\n",
       "       compactness_worst  concavity_worst  concave points_worst  \\\n",
       "count         249.000000       249.000000            249.000000   \n",
       "mean            0.033393         0.029157              0.058714   \n",
       "std             0.086725         0.116995              0.092291   \n",
       "min             0.000171         0.000425              0.001129   \n",
       "25%             0.002153         0.002642              0.001624   \n",
       "50%             0.004059         0.003863              0.002114   \n",
       "75%             0.006248         0.005880              0.182644   \n",
       "max             0.336108         0.572187              0.245207   \n",
       "\n",
       "       symmetry_worst  fractal_dimension_worst  \n",
       "count      249.000000               249.000000  \n",
       "mean         0.040365                 0.009861  \n",
       "std          0.151395                 0.028081  \n",
       "min          0.006040                 0.000906  \n",
       "25%          0.007367                 0.001286  \n",
       "50%          0.009598                 0.007429  \n",
       "75%          0.011780                 0.008121  \n",
       "max          0.778905                 0.173421  \n",
       "\n",
       "[8 rows x 30 columns]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features_train_ros[label_train_smote == \"M\"].describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Al revisar las medidas de tendencia central podemos darnos cuenta que no cambian respecto a las originales."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "                   intercept_scaling=1, l1_ratio=None, max_iter=100,\n",
       "                   multi_class='auto', n_jobs=None, penalty='l2',\n",
       "                   random_state=11, solver='lbfgs', tol=0.0001, verbose=0,\n",
       "                   warm_start=False)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr = sklearn.linear_model.LogisticRegression(random_state=11)\n",
    "lr.fit(features_train_smote, label_train_smote)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr_predictions = lr.predict(features_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           B       0.99      0.96      0.98       108\n",
      "           M       0.67      0.89      0.76         9\n",
      "\n",
      "    accuracy                           0.96       117\n",
      "   macro avg       0.83      0.93      0.87       117\n",
      "weighted avg       0.97      0.96      0.96       117\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(sklearn.metrics.classification_report(label_test, lr_predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Con esta técnica, obtenermos resultados similares con el sobremuestreo de la clase minoritaria."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## _Synthetic Minority Oversampling Technique_ + Submuestro de la clase mayoritaria + Balanceo parcial del conjunto de datos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Los autores de SMOTE recomiendan realizar paralelamente un submuestreo de la clase mayoritaria para poder obtener mejores resultados"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "sm = imblearn.over_sampling.SMOTE(sampling_strategy=0.3,random_state=11) # Hacemos SMOTe hasta obtener un balance de 30% de la clase minoritaria\n",
    "rus = imblearn.under_sampling.RandomUnderSampler(sampling_strategy=0.5,random_state=11) # Hacemos submuestreo hasta obtener un balance del 50%\n",
    "steps = [('SMOTE', sm), ('RUS', rus)] # Ponemos ambos pasos\n",
    "sm_rus = imblearn.pipeline.Pipeline(steps=steps) # Creamos un pipeline para realizar ambas tareas\n",
    "features_train_sm_rus, label_train_sm_rus = sm_rus.fit_resample(features_train, label_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "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.scatter(features_train_sm_rus[label_train_sm_rus == \"B\"].area_worst, \n",
    "            features_train_sm_rus[label_train_sm_rus == \"B\"].perimeter_worst, \n",
    "            label=\"B\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.scatter(features_train_sm_rus[label_train_sm_rus == \"M\"].area_worst, \n",
    "            features_train_sm_rus[label_train_sm_rus == \"M\"].perimeter_worst, \n",
    "            label=\"M\",\n",
    "            alpha=0.5\n",
    "           )\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "                   intercept_scaling=1, l1_ratio=None, max_iter=100,\n",
       "                   multi_class='auto', n_jobs=None, penalty='l2',\n",
       "                   random_state=11, solver='lbfgs', tol=0.0001, verbose=0,\n",
       "                   warm_start=False)"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr = sklearn.linear_model.LogisticRegression(random_state=11)\n",
    "lr.fit(features_train_sm_rus, label_train_sm_rus)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr_predictions = lr.predict(features_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           B       0.99      0.98      0.99       108\n",
      "           M       0.80      0.89      0.84         9\n",
      "\n",
      "    accuracy                           0.97       117\n",
      "   macro avg       0.90      0.94      0.91       117\n",
      "weighted avg       0.98      0.97      0.97       117\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(sklearn.metrics.classification_report(label_test, lr_predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pudimos mejorar las métricas y esta técnica es la que mejor se comporta"
   ]
  },
  {
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}