Artigo - FeatureUnion com Scikit-learn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ** Minerando Dados** -  [www.minerandodados.com.br](http://www.minerandodados.com.br)\n",
    "\n",
    "_Aprenda Data Science e Alavanque sua Carreira_\n",
    "\n",
    "**Autor: Rodrigo Santana**\n",
    "\n",
    "**e-mail: contato@minerandodados.com.br**\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Artigo FeatureUnion"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import cross_val_predict\n",
    "from sklearn import metrics\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import svm\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Ler o dataset.\n",
    "dataset = pd.read_csv('data.csv', sep=',')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "__Separa dados de treino e classes__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Separa a classe dos dados\n",
    "classes = dataset['target']\n",
    "dataset.drop('target', axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "__Pre-processamento de dados__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def remove_features(lista_features):\n",
    "    for i in lista_features:\n",
    "        dataset.drop(i, axis=1, inplace=True)\n",
    "    return 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Remove colunas 'id' e 'title'\n",
    "remove_features(['id','song_title'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "__Aplica o LabelEncoder na coluna 'artist'__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Label Encoder\n",
    "\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "enc = LabelEncoder()\n",
    "inteiros = enc.fit_transform(dataset['artist'])\n",
    "\n",
    "# Cria uma nova coluna chamada 'artist_inteiros'\n",
    "\n",
    "dataset['artist_inteiros'] = inteiros\n",
    "remove_features(['artist'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "__Visualizando os dados__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>acousticness</th>\n",
       "      <th>danceability</th>\n",
       "      <th>duration_ms</th>\n",
       "      <th>energy</th>\n",
       "      <th>instrumentalness</th>\n",
       "      <th>key</th>\n",
       "      <th>liveness</th>\n",
       "      <th>loudness</th>\n",
       "      <th>mode</th>\n",
       "      <th>speechiness</th>\n",
       "      <th>tempo</th>\n",
       "      <th>time_signature</th>\n",
       "      <th>valence</th>\n",
       "      <th>artist_inteiros</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0102</td>\n",
       "      <td>0.833</td>\n",
       "      <td>204600</td>\n",
       "      <td>0.434</td>\n",
       "      <td>0.021900</td>\n",
       "      <td>2</td>\n",
       "      <td>0.1650</td>\n",
       "      <td>-8.795</td>\n",
       "      <td>1</td>\n",
       "      <td>0.4310</td>\n",
       "      <td>150.062</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.286</td>\n",
       "      <td>449</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.1990</td>\n",
       "      <td>0.743</td>\n",
       "      <td>326933</td>\n",
       "      <td>0.359</td>\n",
       "      <td>0.006110</td>\n",
       "      <td>1</td>\n",
       "      <td>0.1370</td>\n",
       "      <td>-10.401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0794</td>\n",
       "      <td>160.083</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.588</td>\n",
       "      <td>222</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0344</td>\n",
       "      <td>0.838</td>\n",
       "      <td>185707</td>\n",
       "      <td>0.412</td>\n",
       "      <td>0.000234</td>\n",
       "      <td>2</td>\n",
       "      <td>0.1590</td>\n",
       "      <td>-7.148</td>\n",
       "      <td>1</td>\n",
       "      <td>0.2890</td>\n",
       "      <td>75.044</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.173</td>\n",
       "      <td>449</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.6040</td>\n",
       "      <td>0.494</td>\n",
       "      <td>199413</td>\n",
       "      <td>0.338</td>\n",
       "      <td>0.510000</td>\n",
       "      <td>5</td>\n",
       "      <td>0.0922</td>\n",
       "      <td>-15.236</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0261</td>\n",
       "      <td>86.468</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.230</td>\n",
       "      <td>95</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.1800</td>\n",
       "      <td>0.678</td>\n",
       "      <td>392893</td>\n",
       "      <td>0.561</td>\n",
       "      <td>0.512000</td>\n",
       "      <td>5</td>\n",
       "      <td>0.4390</td>\n",
       "      <td>-11.648</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0694</td>\n",
       "      <td>174.004</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.904</td>\n",
       "      <td>628</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   acousticness  danceability  duration_ms  energy  instrumentalness  key  \\\n",
       "0        0.0102         0.833       204600   0.434          0.021900    2   \n",
       "1        0.1990         0.743       326933   0.359          0.006110    1   \n",
       "2        0.0344         0.838       185707   0.412          0.000234    2   \n",
       "3        0.6040         0.494       199413   0.338          0.510000    5   \n",
       "4        0.1800         0.678       392893   0.561          0.512000    5   \n",
       "\n",
       "   liveness  loudness  mode  speechiness    tempo  time_signature  valence  \\\n",
       "0    0.1650    -8.795     1       0.4310  150.062             4.0    0.286   \n",
       "1    0.1370   -10.401     1       0.0794  160.083             4.0    0.588   \n",
       "2    0.1590    -7.148     1       0.2890   75.044             4.0    0.173   \n",
       "3    0.0922   -15.236     1       0.0261   86.468             4.0    0.230   \n",
       "4    0.4390   -11.648     0       0.0694  174.004             4.0    0.904   \n",
       "\n",
       "   artist_inteiros  \n",
       "0              449  \n",
       "1              222  \n",
       "2              449  \n",
       "3               95  \n",
       "4              628  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    14657.103436\n",
       "1    23379.485251\n",
       "2    13302.342974\n",
       "3    14256.537593\n",
       "4    28121.121386\n",
       "5    17984.451628\n",
       "6    17273.765893\n",
       "7    25048.888093\n",
       "8    14522.720793\n",
       "9    16247.345506\n",
       "dtype: float64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Visualizando o valor das médias das 10 primeiras linhas.\n",
    "dataset.head(10).mean(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0     54669.464606\n",
       "1     87368.739744\n",
       "2     49621.617927\n",
       "3     53291.706129\n",
       "4    104988.706256\n",
       "5     67162.400181\n",
       "6     64508.038528\n",
       "7     93431.806459\n",
       "8     54205.256516\n",
       "9     60612.891926\n",
       "dtype: float64"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Visualizando o valor de desvio padrão das 10 primeiras linhas.\n",
    "dataset.head(10).std(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "14"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Visualizando o número de colunas\n",
    "len(dataset.columns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "__Classe Stats__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.base import BaseEstimator, TransformerMixin\n",
    "\n",
    "class Stats(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self):\n",
    "        pass\n",
    "    \n",
    "    def transform(self, df, y=None):\n",
    "        df = df.assign(mean=df.mean(axis=1),std=df.std(axis=1))\n",
    "        return df\n",
    "\n",
    "    def fit(self, df, y=None):\n",
    "        return self"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "__ Aplica FeatureUnion __"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# importa a função do PCA e o utilitário FeatureUnion.\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.pipeline import FeatureUnion\n",
    "\n",
    "# Define o número de componentes do PCA.\n",
    "pca = PCA(n_components=1)\n",
    "\n",
    "# Cria um objeto com o PCA e a classe Stats usando o FeatureUnion\n",
    "features = FeatureUnion([(\"pca\" ,pca),\n",
    "                         ('stats',Stats())\n",
    "                        ])\n",
    "# Aplica a combinação criada acima nos dados.\n",
    "dataset = features.fit(dataset,classes).transform(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2017, 17)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Visualizando as Colunas (linhas,colunas)\n",
    "dataset.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>10</th>\n",
       "      <th>11</th>\n",
       "      <th>12</th>\n",
       "      <th>13</th>\n",
       "      <th>14</th>\n",
       "      <th>15</th>\n",
       "      <th>16</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-41706.163219</td>\n",
       "      <td>0.0102</td>\n",
       "      <td>0.833</td>\n",
       "      <td>204600.0</td>\n",
       "      <td>0.434</td>\n",
       "      <td>0.021900</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.1650</td>\n",
       "      <td>-8.795</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.4310</td>\n",
       "      <td>150.062</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.286</td>\n",
       "      <td>449.0</td>\n",
       "      <td>14657.103436</td>\n",
       "      <td>54669.464606</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>80626.870843</td>\n",
       "      <td>0.1990</td>\n",
       "      <td>0.743</td>\n",
       "      <td>326933.0</td>\n",
       "      <td>0.359</td>\n",
       "      <td>0.006110</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.1370</td>\n",
       "      <td>-10.401</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0794</td>\n",
       "      <td>160.083</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.588</td>\n",
       "      <td>222.0</td>\n",
       "      <td>23379.485251</td>\n",
       "      <td>87368.739744</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-60599.161129</td>\n",
       "      <td>0.0344</td>\n",
       "      <td>0.838</td>\n",
       "      <td>185707.0</td>\n",
       "      <td>0.412</td>\n",
       "      <td>0.000234</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.1590</td>\n",
       "      <td>-7.148</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.2890</td>\n",
       "      <td>75.044</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.173</td>\n",
       "      <td>449.0</td>\n",
       "      <td>13302.342974</td>\n",
       "      <td>49621.617927</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-46893.105551</td>\n",
       "      <td>0.6040</td>\n",
       "      <td>0.494</td>\n",
       "      <td>199413.0</td>\n",
       "      <td>0.338</td>\n",
       "      <td>0.510000</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.0922</td>\n",
       "      <td>-15.236</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0261</td>\n",
       "      <td>86.468</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.230</td>\n",
       "      <td>95.0</td>\n",
       "      <td>14256.537593</td>\n",
       "      <td>53291.706129</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>146586.805519</td>\n",
       "      <td>0.1800</td>\n",
       "      <td>0.678</td>\n",
       "      <td>392893.0</td>\n",
       "      <td>0.561</td>\n",
       "      <td>0.512000</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.4390</td>\n",
       "      <td>-11.648</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0694</td>\n",
       "      <td>174.004</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.904</td>\n",
       "      <td>628.0</td>\n",
       "      <td>28121.121386</td>\n",
       "      <td>104988.706256</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              0       1      2         3      4         5    6       7   \\\n",
       "0  -41706.163219  0.0102  0.833  204600.0  0.434  0.021900  2.0  0.1650   \n",
       "1   80626.870843  0.1990  0.743  326933.0  0.359  0.006110  1.0  0.1370   \n",
       "2  -60599.161129  0.0344  0.838  185707.0  0.412  0.000234  2.0  0.1590   \n",
       "3  -46893.105551  0.6040  0.494  199413.0  0.338  0.510000  5.0  0.0922   \n",
       "4  146586.805519  0.1800  0.678  392893.0  0.561  0.512000  5.0  0.4390   \n",
       "\n",
       "       8    9       10       11   12     13     14            15  \\\n",
       "0  -8.795  1.0  0.4310  150.062  4.0  0.286  449.0  14657.103436   \n",
       "1 -10.401  1.0  0.0794  160.083  4.0  0.588  222.0  23379.485251   \n",
       "2  -7.148  1.0  0.2890   75.044  4.0  0.173  449.0  13302.342974   \n",
       "3 -15.236  1.0  0.0261   86.468  4.0  0.230   95.0  14256.537593   \n",
       "4 -11.648  0.0  0.0694  174.004  4.0  0.904  628.0  28121.121386   \n",
       "\n",
       "              16  \n",
       "0   54669.464606  \n",
       "1   87368.739744  \n",
       "2   49621.617927  \n",
       "3   53291.706129  \n",
       "4  104988.706256  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Transforma em Dataframe para melhor visualização\n",
    "dataset = pd.DataFrame(dataset)\n",
    "dataset.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "__Cria um Pipeline para automatizar todos os passos __"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Instancia o PCA.\n",
    "pca = PCA(n_components=1)\n",
    "\n",
    "# Cria um pipeline\n",
    "pipeline = Pipeline([\n",
    "                    ('Features', FeatureUnion([\n",
    "                        ('Stats', Stats()),\n",
    "                        ('PCA', pca)\n",
    "                        ])\n",
    "            ),\n",
    "            ('Scaler', StandardScaler()),      # Aplica um Scaler nos dados.\n",
    "            ('SVM', svm.SVC())                 # Executa um Classificador usando o SVM.\n",
    "         ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "17"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Quantidade de colunas\n",
    "len(dataset.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Features', FeatureUnion(n_jobs=1,\n",
       "         transformer_list=[('Stats', Stats()), ('PCA', PCA(copy=True, iterated_power='auto', n_components=1, random_state=None,\n",
       "    svd_solver='auto', tol=0.0, whiten=False))],\n",
       "         transformer_weights=None)),\n",
       " ('Scaler', StandardScaler(copy=True, with_mean=True, with_std=True)),\n",
       " ('SVM', SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,\n",
       "    decision_function_shape='ovr', degree=3, gamma='auto', kernel='rbf',\n",
       "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
       "    tol=0.001, verbose=False))]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Passos do Pipeline\n",
    "pipeline.steps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Função que retorna a acurácia\n",
    "def Acuracia(clf,X,y):\n",
    "    resultados = cross_val_predict(clf, X, y, cv=5)\n",
    "    return metrics.accuracy_score(y,resultados)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6668319286068418"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Chama a função 'Acuracia'\n",
    "Acuracia(pipeline,dataset,classes)"
   ]
  }
 ],
 "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}