Regressão para previsão de notas de matemática do ENEM 2016.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# DESAFIO CODENATION - NOTAS DO ENEM 2016"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Esse trabalho refere-se a uma tarefa que tive que realizar no desafio da Codenation, para poder participar da Aceleração de Data Science da mesma. No desafio deveríamos usar as notas de alunos (ciências da natureza, linguagens e códigos, ciências humanas e redação) e informações sociais dos alunos (do ano de 2016) que estavam em dois arquivo :train.csv e test.csv.\n",
    "\n",
    "O arquivo train.csv tinha todas as features e com esse arquivo iríamos, após ser tratado, treinar um modelo de Machine Learning do tipo regressão.\n",
    "\n",
    "Já no arquivo train.csv constava todas as features com exceção da feature **NU_NOTA_MT**. Ou seja, após treinarmos o modelo usaríamos as features do arquivo de teste para prever as notas de matemática dos alunos.\n",
    "\n",
    "Após gerarmos as previsões com o modelo de machine learning criaríamos um arquivo com apenas duas colunas que seriam **NU_INSCRICAO** e **NU_NOTA_MT** criaríamos o arquivo csv e o para submeteríamos."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**IMPORTANDO BIBLIOTECAS PARA TRATAMENTO DOS DADOS**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**IMPORTANDO DADOS DE TREINO E TESTE**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "enem_treino=pd.read_csv(\"train.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "enem_teste=pd.read_csv(\"test.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**OBSERVANDO COLUNAS DE CADA BASE DE DADOS**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Unnamed: 0', 'NU_INSCRICAO', 'NU_ANO', 'CO_MUNICIPIO_RESIDENCIA',\n",
       "       'NO_MUNICIPIO_RESIDENCIA', 'CO_UF_RESIDENCIA', 'SG_UF_RESIDENCIA',\n",
       "       'NU_IDADE', 'TP_SEXO', 'TP_ESTADO_CIVIL', 'TP_COR_RACA',\n",
       "       'TP_NACIONALIDADE', 'CO_MUNICIPIO_NASCIMENTO',\n",
       "       'NO_MUNICIPIO_NASCIMENTO', 'CO_UF_NASCIMENTO', 'SG_UF_NASCIMENTO',\n",
       "       'TP_ST_CONCLUSAO', 'TP_ANO_CONCLUIU', 'TP_ESCOLA', 'TP_ENSINO',\n",
       "       'IN_TREINEIRO', 'CO_ESCOLA', 'CO_MUNICIPIO_ESC',\n",
       "       'NO_MUNICIPIO_ESC', 'CO_UF_ESC', 'SG_UF_ESC',\n",
       "       'TP_DEPENDENCIA_ADM_ESC', 'TP_LOCALIZACAO_ESC', 'TP_SIT_FUNC_ESC',\n",
       "       'IN_BAIXA_VISAO', 'IN_CEGUEIRA', 'IN_SURDEZ',\n",
       "       'IN_DEFICIENCIA_AUDITIVA', 'IN_SURDO_CEGUEIRA',\n",
       "       'IN_DEFICIENCIA_FISICA', 'IN_DEFICIENCIA_MENTAL',\n",
       "       'IN_DEFICIT_ATENCAO', 'IN_DISLEXIA', 'IN_DISCALCULIA',\n",
       "       'IN_AUTISMO', 'IN_VISAO_MONOCULAR', 'IN_OUTRA_DEF', 'IN_SABATISTA',\n",
       "       'IN_GESTANTE', 'IN_LACTANTE', 'IN_IDOSO',\n",
       "       'IN_ESTUDA_CLASSE_HOSPITALAR', 'IN_SEM_RECURSO', 'IN_BRAILLE',\n",
       "       'IN_AMPLIADA_24', 'IN_AMPLIADA_18', 'IN_LEDOR', 'IN_ACESSO',\n",
       "       'IN_TRANSCRICAO', 'IN_LIBRAS', 'IN_LEITURA_LABIAL',\n",
       "       'IN_MESA_CADEIRA_RODAS', 'IN_MESA_CADEIRA_SEPARADA',\n",
       "       'IN_APOIO_PERNA', 'IN_GUIA_INTERPRETE', 'IN_MACA', 'IN_COMPUTADOR',\n",
       "       'IN_CADEIRA_ESPECIAL', 'IN_CADEIRA_CANHOTO',\n",
       "       'IN_CADEIRA_ACOLCHOADA', 'IN_PROVA_DEITADO', 'IN_MOBILIARIO_OBESO',\n",
       "       'IN_LAMINA_OVERLAY', 'IN_PROTETOR_AURICULAR', 'IN_MEDIDOR_GLICOSE',\n",
       "       'IN_MAQUINA_BRAILE', 'IN_SOROBAN', 'IN_MARCA_PASSO', 'IN_SONDA',\n",
       "       'IN_MEDICAMENTOS', 'IN_SALA_INDIVIDUAL', 'IN_SALA_ESPECIAL',\n",
       "       'IN_SALA_ACOMPANHANTE', 'IN_MOBILIARIO_ESPECIFICO',\n",
       "       'IN_MATERIAL_ESPECIFICO', 'IN_NOME_SOCIAL', 'IN_CERTIFICADO',\n",
       "       'NO_ENTIDADE_CERTIFICACAO', 'CO_UF_ENTIDADE_CERTIFICACAO',\n",
       "       'SG_UF_ENTIDADE_CERTIFICACAO', 'CO_MUNICIPIO_PROVA',\n",
       "       'NO_MUNICIPIO_PROVA', 'CO_UF_PROVA', 'SG_UF_PROVA',\n",
       "       'TP_PRESENCA_CN', 'TP_PRESENCA_CH', 'TP_PRESENCA_LC',\n",
       "       'TP_PRESENCA_MT', 'CO_PROVA_CN', 'CO_PROVA_CH', 'CO_PROVA_LC',\n",
       "       'CO_PROVA_MT', 'NU_NOTA_CN', 'NU_NOTA_CH', 'NU_NOTA_LC',\n",
       "       'NU_NOTA_MT', 'TX_RESPOSTAS_CN', 'TX_RESPOSTAS_CH',\n",
       "       'TX_RESPOSTAS_LC', 'TX_RESPOSTAS_MT', 'TP_LINGUA',\n",
       "       'TX_GABARITO_CN', 'TX_GABARITO_CH', 'TX_GABARITO_LC',\n",
       "       'TX_GABARITO_MT', 'TP_STATUS_REDACAO', 'NU_NOTA_COMP1',\n",
       "       'NU_NOTA_COMP2', 'NU_NOTA_COMP3', 'NU_NOTA_COMP4', 'NU_NOTA_COMP5',\n",
       "       'NU_NOTA_REDACAO', 'Q001', 'Q002', 'Q003', 'Q004', 'Q005', 'Q006',\n",
       "       'Q007', 'Q008', 'Q009', 'Q010', 'Q011', 'Q012', 'Q013', 'Q014',\n",
       "       'Q015', 'Q016', 'Q017', 'Q018', 'Q019', 'Q020', 'Q021', 'Q022',\n",
       "       'Q023', 'Q024', 'Q025', 'Q026', 'Q027', 'Q028', 'Q029', 'Q030',\n",
       "       'Q031', 'Q032', 'Q033', 'Q034', 'Q035', 'Q036', 'Q037', 'Q038',\n",
       "       'Q039', 'Q040', 'Q041', 'Q042', 'Q043', 'Q044', 'Q045', 'Q046',\n",
       "       'Q047', 'Q048', 'Q049', 'Q050'], dtype=object)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "enem_treino.columns.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['NU_INSCRICAO', 'CO_UF_RESIDENCIA', 'SG_UF_RESIDENCIA', 'NU_IDADE',\n",
       "       'TP_SEXO', 'TP_COR_RACA', 'TP_NACIONALIDADE', 'TP_ST_CONCLUSAO',\n",
       "       'TP_ANO_CONCLUIU', 'TP_ESCOLA', 'TP_ENSINO', 'IN_TREINEIRO',\n",
       "       'TP_DEPENDENCIA_ADM_ESC', 'IN_BAIXA_VISAO', 'IN_CEGUEIRA',\n",
       "       'IN_SURDEZ', 'IN_DISLEXIA', 'IN_DISCALCULIA', 'IN_SABATISTA',\n",
       "       'IN_GESTANTE', 'IN_IDOSO', 'TP_PRESENCA_CN', 'TP_PRESENCA_CH',\n",
       "       'TP_PRESENCA_LC', 'CO_PROVA_CN', 'CO_PROVA_CH', 'CO_PROVA_LC',\n",
       "       'CO_PROVA_MT', 'NU_NOTA_CN', 'NU_NOTA_CH', 'NU_NOTA_LC',\n",
       "       'TP_LINGUA', 'TP_STATUS_REDACAO', 'NU_NOTA_COMP1', 'NU_NOTA_COMP2',\n",
       "       'NU_NOTA_COMP3', 'NU_NOTA_COMP4', 'NU_NOTA_COMP5',\n",
       "       'NU_NOTA_REDACAO', 'Q001', 'Q002', 'Q006', 'Q024', 'Q025', 'Q026',\n",
       "       'Q027', 'Q047'], dtype=object)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "enem_teste.columns.values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Vemos que o número de colunas de treino e teste não são as mesma. Logo iremos excluir colunas da base de treino para ter o mesmo número de colunas. A **'NU_NOTA_MT'** não será excluída, pois será usada para treinar o algoritmo."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#vamos criar uma lista com o nome das colunas da base teste e com a variável 'NU_NOTA_MT'\n",
    "\n",
    "feature=['NU_INSCRICAO', 'CO_UF_RESIDENCIA', 'SG_UF_RESIDENCIA', 'NU_IDADE',\n",
    "       'TP_SEXO', 'TP_COR_RACA', 'TP_NACIONALIDADE', 'TP_ST_CONCLUSAO',\n",
    "       'TP_ANO_CONCLUIU', 'TP_ESCOLA', 'TP_ENSINO', 'IN_TREINEIRO',\n",
    "       'TP_DEPENDENCIA_ADM_ESC', 'IN_BAIXA_VISAO', 'IN_CEGUEIRA',\n",
    "       'IN_SURDEZ', 'IN_DISLEXIA', 'IN_DISCALCULIA', 'IN_SABATISTA',\n",
    "       'IN_GESTANTE', 'IN_IDOSO', 'TP_PRESENCA_CN', 'TP_PRESENCA_CH',\n",
    "       'TP_PRESENCA_LC', 'CO_PROVA_CN', 'CO_PROVA_CH', 'CO_PROVA_LC',\n",
    "       'CO_PROVA_MT', 'NU_NOTA_CN', 'NU_NOTA_CH', 'NU_NOTA_LC',\n",
    "       'TP_LINGUA', 'TP_STATUS_REDACAO', 'NU_NOTA_COMP1', 'NU_NOTA_COMP2',\n",
    "       'NU_NOTA_COMP3', 'NU_NOTA_COMP4', 'NU_NOTA_COMP5',\n",
    "       'NU_NOTA_REDACAO', 'Q001', 'Q002', 'Q006', 'Q024', 'Q025', 'Q026',\n",
    "       'Q027', 'Q047','NU_NOTA_MT']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Agora vamos definí-las como colunas do dataset de treino\n",
    "\n",
    "enem_treino=enem_treino[feature]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['NU_INSCRICAO', 'CO_UF_RESIDENCIA', 'SG_UF_RESIDENCIA', 'NU_IDADE',\n",
       "       'TP_SEXO', 'TP_COR_RACA', 'TP_NACIONALIDADE', 'TP_ST_CONCLUSAO',\n",
       "       'TP_ANO_CONCLUIU', 'TP_ESCOLA', 'TP_ENSINO', 'IN_TREINEIRO',\n",
       "       'TP_DEPENDENCIA_ADM_ESC', 'IN_BAIXA_VISAO', 'IN_CEGUEIRA',\n",
       "       'IN_SURDEZ', 'IN_DISLEXIA', 'IN_DISCALCULIA', 'IN_SABATISTA',\n",
       "       'IN_GESTANTE', 'IN_IDOSO', 'TP_PRESENCA_CN', 'TP_PRESENCA_CH',\n",
       "       'TP_PRESENCA_LC', 'CO_PROVA_CN', 'CO_PROVA_CH', 'CO_PROVA_LC',\n",
       "       'CO_PROVA_MT', 'NU_NOTA_CN', 'NU_NOTA_CH', 'NU_NOTA_LC',\n",
       "       'TP_LINGUA', 'TP_STATUS_REDACAO', 'NU_NOTA_COMP1', 'NU_NOTA_COMP2',\n",
       "       'NU_NOTA_COMP3', 'NU_NOTA_COMP4', 'NU_NOTA_COMP5',\n",
       "       'NU_NOTA_REDACAO', 'Q001', 'Q002', 'Q006', 'Q024', 'Q025', 'Q026',\n",
       "       'Q027', 'Q047', 'NU_NOTA_MT'], dtype=object)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#visualizando as colunas novamente\n",
    "enem_treino.columns.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['NU_INSCRICAO', 'CO_UF_RESIDENCIA', 'SG_UF_RESIDENCIA', 'NU_IDADE',\n",
       "       'TP_SEXO', 'TP_COR_RACA', 'TP_NACIONALIDADE', 'TP_ST_CONCLUSAO',\n",
       "       'TP_ANO_CONCLUIU', 'TP_ESCOLA', 'TP_ENSINO', 'IN_TREINEIRO',\n",
       "       'TP_DEPENDENCIA_ADM_ESC', 'IN_BAIXA_VISAO', 'IN_CEGUEIRA',\n",
       "       'IN_SURDEZ', 'IN_DISLEXIA', 'IN_DISCALCULIA', 'IN_SABATISTA',\n",
       "       'IN_GESTANTE', 'IN_IDOSO', 'TP_PRESENCA_CN', 'TP_PRESENCA_CH',\n",
       "       'TP_PRESENCA_LC', 'CO_PROVA_CN', 'CO_PROVA_CH', 'CO_PROVA_LC',\n",
       "       'CO_PROVA_MT', 'NU_NOTA_CN', 'NU_NOTA_CH', 'NU_NOTA_LC',\n",
       "       'TP_LINGUA', 'TP_STATUS_REDACAO', 'NU_NOTA_COMP1', 'NU_NOTA_COMP2',\n",
       "       'NU_NOTA_COMP3', 'NU_NOTA_COMP4', 'NU_NOTA_COMP5',\n",
       "       'NU_NOTA_REDACAO', 'Q001', 'Q002', 'Q006', 'Q024', 'Q025', 'Q026',\n",
       "       'Q027', 'Q047'], dtype=object)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "enem_teste.columns.values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pronto! Como podemos observar, agora a única diferença é que a base de treino tem a variável NU_NOTA_MT.\n",
    "\n",
    "Agora quais features ou variáveis usar ?\n",
    "\n",
    "Uma bom modo de definir as variáveis para previsão é observando a correlação entre as variáveis e a identificar as que possuem maior correlação com a variável a ser predita."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "#vamos importar a biblioteca seaborn e matplotlib para plotar o gráfico de correlações\n",
    "\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CO_UF_RESIDENCIA          0.107493\n",
      "NU_IDADE                 -0.061485\n",
      "TP_COR_RACA              -0.157364\n",
      "TP_NACIONALIDADE         -0.034965\n",
      "TP_ST_CONCLUSAO          -0.087869\n",
      "TP_ANO_CONCLUIU           0.029039\n",
      "TP_ESCOLA                 0.069304\n",
      "TP_ENSINO                -0.086353\n",
      "IN_TREINEIRO             -0.022287\n",
      "TP_DEPENDENCIA_ADM_ESC    0.373233\n",
      "IN_BAIXA_VISAO            0.000685\n",
      "IN_CEGUEIRA                    NaN\n",
      "IN_SURDEZ                -0.015249\n",
      "IN_DISLEXIA               0.001453\n",
      "IN_DISCALCULIA            0.009813\n",
      "IN_SABATISTA             -0.016459\n",
      "IN_GESTANTE              -0.000945\n",
      "IN_IDOSO                 -0.011336\n",
      "TP_PRESENCA_CN            0.003552\n",
      "TP_PRESENCA_CH            0.003552\n",
      "TP_PRESENCA_LC                 NaN\n",
      "NU_NOTA_CN                0.584941\n",
      "NU_NOTA_CH                0.529594\n",
      "NU_NOTA_LC                0.494695\n",
      "TP_LINGUA                -0.201726\n",
      "TP_STATUS_REDACAO        -0.059376\n",
      "NU_NOTA_COMP1             0.299402\n",
      "NU_NOTA_COMP2             0.335638\n",
      "NU_NOTA_COMP3             0.350307\n",
      "NU_NOTA_COMP4             0.342282\n",
      "NU_NOTA_COMP5             0.343337\n",
      "NU_NOTA_REDACAO           0.379376\n",
      "NU_NOTA_MT                1.000000\n",
      "Name: NU_NOTA_MT, dtype: float64\n"
     ]
    }
   ],
   "source": [
    "#Vamos atribuir uma variável a correalção entre as features\n",
    "correlacao=enem_treino.corr()\n",
    "\n",
    "#Vamos agora imprimir a correlação das features com 'NU_NOTA_MT'\n",
    "print(correlacao['NU_NOTA_MT'])\n",
    "\n",
    "#Vemos abaixo que as variáveis 'TP_COR_RACA','NU_NOTA_CN','NU_NOTA_CH','NU_NOTA_LC','TP_LINGUA','NU_NOTA_COMP1',\n",
    "#                              'NU_NOTA_COMP2','NU_NOTA_COMP3','NU_NOTA_COMP4','NU_NOTA_COMP5','NU_NOTA_REDACAO'\n",
    "#são as que possuem maiores resultado de correlação com 'NU_NOTA_MT'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Vamos agora selecionar apenas essas variáveis apenas essas variáveis para as bases de treino e teste"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "features1=['TP_COR_RACA','NU_NOTA_CN','NU_NOTA_CH','NU_NOTA_LC',\n",
    "           'TP_LINGUA','NU_NOTA_COMP1','NU_NOTA_COMP2','NU_NOTA_COMP3','NU_NOTA_COMP4','NU_NOTA_COMP5',\n",
    "           'NU_NOTA_REDACAO','NU_NOTA_MT']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "features2=['TP_COR_RACA','NU_NOTA_CN','NU_NOTA_CH','NU_NOTA_LC',\n",
    "           'TP_LINGUA','NU_NOTA_COMP1','NU_NOTA_COMP2','NU_NOTA_COMP3','NU_NOTA_COMP4','NU_NOTA_COMP5',\n",
    "           'NU_NOTA_REDACAO']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "enem_treino=enem_treino[features1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "enem_teste=enem_teste[features2]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**VISUALIZANDO NOVAMENTE AS COLUNAS DE CADA DATASET**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['TP_COR_RACA', 'NU_NOTA_CN', 'NU_NOTA_CH', 'NU_NOTA_LC',\n",
       "       'TP_LINGUA', 'NU_NOTA_COMP1', 'NU_NOTA_COMP2', 'NU_NOTA_COMP3',\n",
       "       'NU_NOTA_COMP4', 'NU_NOTA_COMP5', 'NU_NOTA_REDACAO', 'NU_NOTA_MT'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "enem_treino.columns.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['TP_COR_RACA', 'NU_NOTA_CN', 'NU_NOTA_CH', 'NU_NOTA_LC',\n",
       "       'TP_LINGUA', 'NU_NOTA_COMP1', 'NU_NOTA_COMP2', 'NU_NOTA_COMP3',\n",
       "       'NU_NOTA_COMP4', 'NU_NOTA_COMP5', 'NU_NOTA_REDACAO'], dtype=object)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "enem_teste.columns.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x183d10b1cc0>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Vamos ver novamente a correlação entre as features\n",
    "\n",
    "%matplotlib inline\n",
    "correlacao2=enem_treino.corr()\n",
    "sns.heatmap(correlacao2, annot = True, cmap = \"YlOrRd\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**OBSERVANDO MISSING VALUES DOS CONJUNTOS DE DADOS**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TP_COR_RACA           0\n",
       "NU_NOTA_CN         3389\n",
       "NU_NOTA_CH         3389\n",
       "NU_NOTA_LC         3597\n",
       "TP_LINGUA             0\n",
       "NU_NOTA_COMP1      3597\n",
       "NU_NOTA_COMP2      3597\n",
       "NU_NOTA_COMP3      3597\n",
       "NU_NOTA_COMP4      3597\n",
       "NU_NOTA_COMP5      3597\n",
       "NU_NOTA_REDACAO    3597\n",
       "NU_NOTA_MT         3597\n",
       "dtype: int64"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "enem_treino.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TP_COR_RACA           0\n",
       "NU_NOTA_CN         1134\n",
       "NU_NOTA_CH         1134\n",
       "NU_NOTA_LC         1199\n",
       "TP_LINGUA             0\n",
       "NU_NOTA_COMP1      1199\n",
       "NU_NOTA_COMP2      1199\n",
       "NU_NOTA_COMP3      1199\n",
       "NU_NOTA_COMP4      1199\n",
       "NU_NOTA_COMP5      1199\n",
       "NU_NOTA_REDACAO    1199\n",
       "dtype: int64"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "enem_teste.isnull().sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Como podemos ver nas variáveis acima (tanto na base de treino como na de teste) há muitos valores faltantes. Há duas opções para resolvermos esse problema: exclusão das observações NAN's ou faremos um processo de imputação. Com a exclusão de valores faltantes não houveram bons resultados dos modelos utilizados e não consegui chegar no score mínimo de 90% para passar no desafio da Codenation. Dessa forma segui por outro caminho que foi a imputação dos valores faltantes.\n",
    "\n",
    "No processo de imputação há algumas alternativas que são:\n",
    "\n",
    "**1)** substituir os valores NAN's por uma constante;\n",
    "\n",
    "**2)** Substituir os valores NAN's pela média;\n",
    "\n",
    "**3)** Substituir os valores NAN's pela mediana;\n",
    "\n",
    "**4)** Substituir os valores NAN's pela moda\n",
    "\n",
    "Como os valores faltantes não são discretos, podemos usar as opções 1,2 e 3.\n",
    "\n",
    "Para esse trabalho optei por substituir por uma constante que foi **0 (zero)**."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**SUBSTITUINDO MISSING VALUES POR ZERO**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "enem_treino['NU_NOTA_CN'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_CH'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_LC'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_COMP1'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_COMP2'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_COMP3'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_COMP4'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_COMP5'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_REDACAO'].fillna(0,inplace=True)\n",
    "enem_treino['NU_NOTA_MT'].fillna(0,inplace=True)\n",
    "\n",
    "enem_teste['NU_NOTA_CN'].fillna(0,inplace=True)\n",
    "enem_teste['NU_NOTA_CH'].fillna(0,inplace=True)\n",
    "enem_teste['NU_NOTA_LC'].fillna(0,inplace=True)\n",
    "enem_teste['NU_NOTA_COMP1'].fillna(0,inplace=True)\n",
    "enem_teste['NU_NOTA_COMP2'].fillna(0,inplace=True)\n",
    "enem_teste['NU_NOTA_COMP3'].fillna(0,inplace=True)\n",
    "enem_teste['NU_NOTA_COMP4'].fillna(0,inplace=True)\n",
    "enem_teste['NU_NOTA_COMP5'].fillna(0,inplace=True)\n",
    "enem_teste['NU_NOTA_REDACAO'].fillna(0,inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**SEPARANDO DADOS DE TREINO**\n",
    "\n",
    "Agora separaremos os dados de treino entre as variáveis e a classe a ser predita pelos modelos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "classe_treino=enem_treino['NU_NOTA_MT']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "treino=enem_treino.drop('NU_NOTA_MT',axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(13730, 11) (13730,) (4576, 11)\n"
     ]
    }
   ],
   "source": [
    "print(treino.shape,classe_treino.shape,enem_teste.shape)\n",
    "\n",
    "#Vemos que a base de treino com as features possuem 11 colunas e a de classe 1 totalizando 12 colunas\n",
    "\n",
    "#A base de teste possui 11 colunas, pois não tem a feature 'NU_NOTA_MT' que será predita."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**USANDO STANDARDSCALER**\n",
    "\n",
    "Nessa etapa colocaremos as variáveis na mesma escala. O StandarScaler coloca os dados em uma distribuição Normal-padrão, ou seja, a média 0 e variância 1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "#importando o módulo StandardScaler da Scikit-learn\n",
    "from sklearn.preprocessing import StandardScaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Instanciando o StandardScaler\n",
    "sc = StandardScaler()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0        399.4\n",
       "1        459.8\n",
       "2          0.0\n",
       "3          0.0\n",
       "4          0.0\n",
       "         ...  \n",
       "13725    403.2\n",
       "13726    452.4\n",
       "13727    398.0\n",
       "13728    386.6\n",
       "13729    428.9\n",
       "Name: NU_NOTA_MT, Length: 13730, dtype: float64"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Aplicando o StandardScaler nas bases de treino e teste\n",
    "\n",
    "treino=sc.fit_transform(treino)\n",
    "enem_teste=sc.fit_transform(enem_teste)\n",
    "\n",
    "#Optei por não aplicar o StandardScaler nos dados de 'NU_NOTA_MT'\n",
    "classe_treino"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Enfim, feito todo o pré-processamento de dados irei aplicar os modelos de Machine Learning de regressão.\n",
    "\n",
    "Não usarei apenas um modelo de regressão. Optei por usar vários modelos, após feitas a previsões submeti os resultados de predição dos modelos e obtive os scores de cada modelo."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO LINEAR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Importando o modelo regressão linear\n",
    "\n",
    "from sklearn.linear_model import LinearRegression\n",
    "\n",
    "#instanciando o modelo\n",
    "\n",
    "enem_model=LinearRegression()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None, normalize=False)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#treinando o modelo com os dados de treino\n",
    "\n",
    "enem_model.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([427.23216888, 479.34007381, 571.12204087, ..., 601.56033082,\n",
       "       489.49825414,  -1.15194999])"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#gerando as previões com o modelo\n",
    "\n",
    "enem_model.predict(enem_teste)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Atribuindo a uma variável os valores preditos\n",
    "\n",
    "y_pred_lr=enem_model.predict(enem_teste)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([427.23216888, 479.34007381, 571.12204087, ..., 601.56033082,\n",
       "       489.49825414,  -1.15194999])"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Valores preditos\n",
    "\n",
    "y_pred_lr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "#atribuindo uma lista com variáveis a serem criadas no pdf\n",
    "features4=['NU_INSCRICAO','NU_NOTA_MT']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste2=pd.read_csv(\"test.csv\")\n",
    "enem_teste2['NU_NOTA_MT']=y_pred_lr\n",
    "enem_teste2=enem_teste2[features4]\n",
    "enem_teste2.to_csv('answer.csv',index=False)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão linear obtive o score de **92.72%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SUPPORT VECTOR REGRESSION"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Importando o modelo Support Vector Regression\n",
    "\n",
    "from sklearn import svm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Instanciando e treinando o modelo\n",
    "\n",
    "clf=svm.SVR().fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando as previsões e atribuindo a uma variável\n",
    "\n",
    "y_pred_clf = clf.predict(enem_teste)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([427.33169087, 445.45026649, 470.82015965, ..., 566.19248287,\n",
       "       446.72085644,  -1.34037223])"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Valores preditos\n",
    "\n",
    "y_pred_clf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste1=pd.read_csv(\"test.csv\")\n",
    "enem_teste1['NU_NOTA_MT']=y_pred_clf\n",
    "enem_teste1=enem_teste1[features4]\n",
    "enem_teste1.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de Support Vector Regression obtive o score de **92.36%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO ÁRVORE DE DECISÃO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Importando o modelo de Regressão de Árvore de Decisão\n",
    "\n",
    "from sklearn.tree import DecisionTreeRegressor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Instanciando o modelo\n",
    "\n",
    "enem_dtr=DecisionTreeRegressor(criterion='mae',\n",
    "                               max_depth=8,\n",
    "                               max_leaf_nodes=None,\n",
    "                               min_impurity_split=None,\n",
    "                               min_samples_leaf=1,\n",
    "                               min_samples_split=2,\n",
    "                               min_weight_fraction_leaf=0.0,\n",
    "                               random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DecisionTreeRegressor(ccp_alpha=0.0, criterion='mae', max_depth=8,\n",
       "                      max_features=None, max_leaf_nodes=None,\n",
       "                      min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                      min_samples_leaf=1, min_samples_split=2,\n",
       "                      min_weight_fraction_leaf=0.0, presort='deprecated',\n",
       "                      random_state=0, splitter='best')"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Treinando o modelo\n",
    "\n",
    "enem_dtr.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([416.45, 429.2 , 590.8 , ..., 725.2 , 444.75,   0.  ])"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Gerando as previsões e atribuindo a uma variável\n",
    "\n",
    "y_dtr_pred=enem_dtr.predict(enem_teste)\n",
    "y_dtr_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste3=pd.read_csv(\"test.csv\")\n",
    "enem_teste3['NU_NOTA_MT']=y_dtr_pred\n",
    "enem_teste3=enem_teste3[features4]\n",
    "enem_teste3.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão de árvore de decisão obtive o score de **91.23%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO RIDGE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Importando o modelo de regressão Ridge\n",
    "\n",
    "from sklearn.linear_model import Ridge"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Ridge(alpha=0.05, copy_X=True, fit_intercept=True, max_iter=None,\n",
       "      normalize=False, random_state=None, solver='cholesky', tol=0.001)"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Instanciando e treinando o modelo\n",
    "\n",
    "enem_ridge=Ridge(alpha=0.05,solver='cholesky')\n",
    "enem_ridge.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando as previsões\n",
    "\n",
    "y_rid_pred=enem_ridge.predict(enem_teste)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([427.23111008, 479.3393302 , 571.11688897, ..., 601.55969094,\n",
       "       489.49629913,  -1.15172572])"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Valores preditos\n",
    "\n",
    "y_rid_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste4=pd.read_csv(\"test.csv\")\n",
    "enem_teste4['NU_NOTA_MT']=y_rid_pred\n",
    "enem_teste4=enem_teste4[features4]\n",
    "enem_teste4.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão ridge obtive o score de **92.72%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO LASSO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Lasso(alpha=0.01, copy_X=True, fit_intercept=True, max_iter=1000,\n",
       "      normalize=False, positive=False, precompute=False, random_state=None,\n",
       "      selection='cyclic', tol=0.0001, warm_start=False)"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Importando, instanciando e treinando o modelo\n",
    "\n",
    "from sklearn.linear_model import Lasso\n",
    "enem_lasso=Lasso(alpha=0.01)\n",
    "enem_lasso.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([427.13702724, 479.37363616, 571.07423343, ..., 601.45677408,\n",
       "       489.45812349,  -1.15479482])"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Gerando as previsões\n",
    "\n",
    "y_pred_lasso=enem_lasso.predict(enem_teste)\n",
    "y_pred_lasso"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste5=pd.read_csv(\"test.csv\")\n",
    "enem_teste5['NU_NOTA_MT']=y_pred_lasso\n",
    "enem_teste5=enem_teste5[features4]\n",
    "enem_teste5.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão lasso obtive o score de **92.72%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO BAGGING"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Treinando e instanciando o modelo\n",
    "\n",
    "from sklearn.ensemble import BaggingRegressor\n",
    "enem_bag=BaggingRegressor(max_samples=70)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "BaggingRegressor(base_estimator=None, bootstrap=True, bootstrap_features=False,\n",
       "                 max_features=1.0, max_samples=70, n_estimators=10, n_jobs=None,\n",
       "                 oob_score=False, random_state=None, verbose=0,\n",
       "                 warm_start=False)"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Treinando o modelo\n",
    "\n",
    "enem_bag.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([409.9 , 451.42, 574.62, ..., 675.79, 470.5 ,   0.  ])"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Gerando as previsões\n",
    "\n",
    "y_pred_bag=enem_bag.predict(enem_teste)\n",
    "y_pred_bag"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste6=pd.read_csv(\"test.csv\")\n",
    "enem_teste6['NU_NOTA_MT']=y_pred_bag\n",
    "enem_teste6=enem_teste6[features4]\n",
    "enem_teste6.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão Bagging obtive o score de **93.17%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO ADABOOST"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "AdaBoostRegressor(base_estimator=None, learning_rate=1.0, loss='linear',\n",
       "                  n_estimators=50, random_state=None)"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Importando, instanciando e treinando o modelo\n",
    "\n",
    "from sklearn.ensemble import AdaBoostRegressor\n",
    "enem_ada=AdaBoostRegressor()\n",
    "enem_ada.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([480.72307692, 465.42990937, 585.94277209, ..., 632.40129234,\n",
       "       480.2752387 ,  36.3010101 ])"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Gerando as previsões\n",
    "y_pred_ada=enem_ada.predict(enem_teste)\n",
    "y_pred_ada"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste7=pd.read_csv(\"test.csv\")\n",
    "enem_teste7['NU_NOTA_MT']=y_pred_ada\n",
    "enem_teste7=enem_teste7[features4]\n",
    "enem_teste7.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão Adaboost obtive o score de **89.89%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO GRADIENT BOOSTING"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GradientBoostingRegressor(alpha=0.9, ccp_alpha=0.0, criterion='friedman_mse',\n",
       "                          init=None, learning_rate=0.1, loss='ls', max_depth=3,\n",
       "                          max_features=None, max_leaf_nodes=None,\n",
       "                          min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                          min_samples_leaf=1, min_samples_split=2,\n",
       "                          min_weight_fraction_leaf=0.0, n_estimators=100,\n",
       "                          n_iter_no_change=None, presort='deprecated',\n",
       "                          random_state=None, subsample=1.0, tol=0.0001,\n",
       "                          validation_fraction=0.1, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Importando, instanciando e treinando o modelo\n",
    "\n",
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "enem_gbr=GradientBoostingRegressor()\n",
    "enem_gbr.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([4.35584337e+02, 4.47240068e+02, 5.93311595e+02, ...,\n",
       "       6.79622049e+02, 4.45750053e+02, 4.58602669e-01])"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Gerando as previsões\n",
    "\n",
    "y_pred_gbr=enem_gbr.predict(enem_teste)\n",
    "y_pred_gbr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste8=pd.read_csv(\"test.csv\")\n",
    "enem_teste8['NU_NOTA_MT']=y_pred_gbr\n",
    "enem_teste8=enem_teste8[features4]\n",
    "enem_teste8.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão gradiente boosting obtive o score de **93.63%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO ELASTICNET"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ElasticNet(alpha=0.01, copy_X=True, fit_intercept=True, l1_ratio=0.5,\n",
       "           max_iter=1000, normalize=False, positive=False, precompute=False,\n",
       "           random_state=None, selection='cyclic', tol=0.0001, warm_start=False)"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Importando, instanciando e treinando o modelo\n",
    "\n",
    "from sklearn.linear_model import ElasticNet\n",
    "enem_en=ElasticNet(alpha=0.01,l1_ratio=0.5)\n",
    "enem_en.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([425.99751976, 478.33716784, 564.61607548, ..., 600.8045946 ,\n",
       "       487.08043907,  -0.81814107])"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Gerando as previsões\n",
    "\n",
    "enem_pred_en=enem_en.predict(enem_teste)\n",
    "enem_pred_en"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste9=pd.read_csv(\"test.csv\")\n",
    "enem_teste9['NU_NOTA_MT']=enem_pred_en\n",
    "enem_teste9=enem_teste9[features4]\n",
    "enem_teste9.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão elasticnet obtive o score de **92.72%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO STOCHASTIC GRADIENT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SGDRegressor(alpha=0.05, average=False, early_stopping=False, epsilon=0.1,\n",
       "             eta0=0.01, fit_intercept=True, l1_ratio=0.15,\n",
       "             learning_rate='invscaling', loss='squared_loss', max_iter=1000,\n",
       "             n_iter_no_change=5, penalty='l2', power_t=0.25, random_state=None,\n",
       "             shuffle=True, tol=0.001, validation_fraction=0.1, verbose=0,\n",
       "             warm_start=False)"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Importando, instanciando e treinando o modelo\n",
    "\n",
    "from sklearn.linear_model import SGDRegressor\n",
    "enem_sgd=SGDRegressor(penalty='l2',alpha=0.05)\n",
    "enem_sgd.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([4.22131349e+02, 4.70064124e+02, 5.30735867e+02, ...,\n",
       "       6.00418302e+02, 4.73692291e+02, 3.30911851e-01])"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Gerando as previsões\n",
    "\n",
    "y_pred_sgd=enem_sgd.predict(enem_teste)\n",
    "y_pred_sgd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste10=pd.read_csv(\"test.csv\")\n",
    "enem_teste10['NU_NOTA_MT']=y_pred_sgd\n",
    "enem_teste10=enem_teste10[features4]\n",
    "enem_teste10.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão stochastic gradient obtive o score de **92.61%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REGRESSÃO XGBOOST"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Importando o modelo\n",
    "\n",
    "import xgboost as xgb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Instanciando o modelo\n",
    "\n",
    "enem_xgb = xgb.XGBRegressor(n_estimators=100, learning_rate=0.08, gamma=10, subsample=0.9,\n",
    "                           colsample_bytree=1, max_depth=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "XGBRegressor(base_score=0.5, booster=None, colsample_bylevel=1,\n",
       "             colsample_bynode=1, colsample_bytree=1, gamma=10, gpu_id=-1,\n",
       "             importance_type='gain', interaction_constraints=None,\n",
       "             learning_rate=0.08, max_delta_step=0, max_depth=10,\n",
       "             min_child_weight=1, missing=nan, monotone_constraints=None,\n",
       "             n_estimators=100, n_jobs=0, num_parallel_tree=1,\n",
       "             objective='reg:squarederror', random_state=0, reg_alpha=0,\n",
       "             reg_lambda=1, scale_pos_weight=1, subsample=0.9, tree_method=None,\n",
       "             validate_parameters=False, verbosity=None)"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Treinando o modelo\n",
    "\n",
    "enem_xgb.fit(treino,classe_treino)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando as previões\n",
    "\n",
    "y_pred_xgb=enem_xgb.predict(enem_teste)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Gerando o arquivo csv com a coluna 'NU_INSCRICAO' e com uma coluna 'NU_NOTA_MT'\n",
    "\n",
    "enem_teste11=pd.read_csv(\"test.csv\")\n",
    "enem_teste11['NU_NOTA_MT']=y_pred_xgb\n",
    "enem_teste11=enem_teste11[features4]\n",
    "enem_teste11.to_csv('answer.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para o modelo de regressão XGboost obtive o score de **93.58%**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparando os resultados entre os modelos\n",
    "\n",
    "Abaixo podemos ver uma tabela comparando os score que cada modelo obteve no desfio da Codenation. O modelo com melhor desempenho foi a Regressão Gradient Boosting com um Score de 93.63%, conseguindo assim alcançar o score mínimo de 90% e participar da Aceleração."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "|                                 ||   SCORE     ||\n",
    "|:-------------------------------:||:-----------:||\n",
    "|**REGRESSÃO GRADIENT BOOSTING**      ||   **93.63%**    ||\n",
    "|REGRESSÃO XGBOOST                ||   93.58%    ||\n",
    "|REGRESSÃO BAGGING                ||   93.17%    ||\n",
    "|REGRESSÃO LINEAR                 ||   92.72%    ||\n",
    "|REGRESSÃO ELASTICNET             ||   92.72%    ||\n",
    "|REGRESSÃO RIDGE                  ||   92.72%    ||\n",
    "|REGRESSÃO LASSO                  ||   92.72%    ||\n",
    "|REGRESSÃO STOCHASTIC GRADIENT    ||   92.61%    ||   \n",
    "|SUPPORT VECTOR REGRESSION        ||   92.36%    ||\n",
    "|REGRESSÃO ÁRVORE DE DECISÃO      ||   91.23%    ||\n",
    "|REGRESSÃO ADABOOST               ||   89.89%    ||\n",
    "\n",
    "Obs: alguns modelos estão com o score igual, mas na verdade há diferença de algumas casas decimais."
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}