4-Melhorando o modelo/otimizacao.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para a apresentação final da nossa análise para o Alura Voz, precisamos definir o melhor modelo e utilizar os melhores hiperparametros nele."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Importando bibliotecas\n",
    "\n",
    "Para a aplicação utilizaremos `pandas`, `seaborn`, `sklearn`, `imblearn` e `sys`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.append('../')\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from utils.utils import plot_matriz_confusao\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "SEED = 42"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Otimizando o modelo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sabendo que é preciso escolher o melhor modelo desenvolvido, faremos uma análise do resultado obtido no [estudo anterior](https://github.com/sthemonica/challenge_dados_1/blob/bruno/3-Modelos%20de%20ML/modelos.ipynb), ordenado pelo valor de Recall:\n",
    "\n",
    "<h3>Tabela dos modelos ordenada pelo Recall</h3>\n",
    "\n",
    "|    | Modelo        |   Acurácia Treino |   Acurácia Teste |   Precisão |   Recall |   F1-Score |\n",
    "|---:|:--------------|------------------:|-----------------:|-----------:|---------:|-----------:|\n",
    "|  0 | random forest |          0.809818 |         0.800928 |   0.769395 | 0.849843 |   0.80762  |\n",
    "|  1 | decision tree |          0.814715 |         0.804407 |   0.776335 | 0.845912 |   0.809631 |\n",
    "|  2 | svc           |          0.751965 |         0.750676 |   0.733086 | 0.775157 |   0.753535 |\n",
    "\n",
    "O ***Recall*** é uma métrica que avalia se as previsões positivas foram dadas corretamente, ou seja, avalia a quantidade de Verdadeiros Positivos (True Pos) e Falsos Negativos (False Neg). De forma que, quanto maior o Recall mais acertos o modelo obteve. \n",
    "\n",
    "Dessa forma, utilizaremos o modelo que obteve o melhor Recall, sendo assim, o Random Forest.\n",
    "\n",
    "Tendo escolhido o melhor modelo, utilizaremos o [GridSearchCV](https://cursos.alura.com.br/course/machine-learning-otimizacao-de-modelos-atraves-de-hiperparametros/task/48715) para otimizá-lo."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Agora, vamos ler o conjunto de dados trabalhado no notebook anterior `modelos.ipynb` através do método `read_json` do pandas:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "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>gender</th>\n",
       "      <th>SeniorCitizen</th>\n",
       "      <th>Partner</th>\n",
       "      <th>Dependents</th>\n",
       "      <th>tenure</th>\n",
       "      <th>PhoneService</th>\n",
       "      <th>MultipleLines</th>\n",
       "      <th>OnlineSecurity</th>\n",
       "      <th>OnlineBackup</th>\n",
       "      <th>DeviceProtection</th>\n",
       "      <th>...</th>\n",
       "      <th>PaymentMethod_Credit card (automatic)</th>\n",
       "      <th>PaymentMethod_Electronic check</th>\n",
       "      <th>PaymentMethod_Mailed check</th>\n",
       "      <th>Contract_Month-to-month</th>\n",
       "      <th>Contract_One year</th>\n",
       "      <th>Contract_Two year</th>\n",
       "      <th>InternetService_DSL</th>\n",
       "      <th>InternetService_Fiber optic</th>\n",
       "      <th>InternetService_No</th>\n",
       "      <th>Churn</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>9</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>9</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>13</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 26 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   gender  SeniorCitizen  Partner  Dependents  tenure  PhoneService  \\\n",
       "0       1              0        1           1       9             1   \n",
       "1       0              0        0           0       9             1   \n",
       "2       0              0        0           0       4             1   \n",
       "3       0              1        1           0      13             1   \n",
       "4       1              1        1           0       3             1   \n",
       "\n",
       "   MultipleLines  OnlineSecurity  OnlineBackup  DeviceProtection  ...  \\\n",
       "0              0               0             1                 0  ...   \n",
       "1              1               0             0                 0  ...   \n",
       "2              0               0             0                 1  ...   \n",
       "3              0               0             1                 1  ...   \n",
       "4              0               0             0                 0  ...   \n",
       "\n",
       "   PaymentMethod_Credit card (automatic)  PaymentMethod_Electronic check  \\\n",
       "0                                      0                               0   \n",
       "1                                      0                               0   \n",
       "2                                      0                               1   \n",
       "3                                      0                               1   \n",
       "4                                      0                               0   \n",
       "\n",
       "   PaymentMethod_Mailed check  Contract_Month-to-month  Contract_One year  \\\n",
       "0                           1                        0                  1   \n",
       "1                           1                        1                  0   \n",
       "2                           0                        1                  0   \n",
       "3                           0                        1                  0   \n",
       "4                           1                        1                  0   \n",
       "\n",
       "   Contract_Two year  InternetService_DSL  InternetService_Fiber optic  \\\n",
       "0                  0                    1                            0   \n",
       "1                  0                    1                            0   \n",
       "2                  0                    0                            1   \n",
       "3                  0                    0                            1   \n",
       "4                  0                    0                            1   \n",
       "\n",
       "   InternetService_No  Churn  \n",
       "0                   0      0  \n",
       "1                   0      0  \n",
       "2                   0      1  \n",
       "3                   0      1  \n",
       "4                   0      1  \n",
       "\n",
       "[5 rows x 26 columns]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dados = pd.read_json(\"../Dados/Telco-Customer-Churn-balanceamento.json\")\n",
    "dados.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para otimizar o modelo, primeiro precisamos criar o conjunto de treino de teste. Desse modo, dividimos o conjunto de dados em entrada (X) e saida (y), ou alvo."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = dados.drop(['Churn'], axis=1)\n",
    "y = dados['Churn']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "E assim podemos separar os dados em treino e teste com `train_test_split`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, x_test, y_train, y_test = train_test_split(X, y, random_state=SEED)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "É preciso definir os hiperparâmetros a serem otimizados pelo GridSearchCV, os quais:\n",
    "\n",
    "* **`n_estimators`**: número de árvores de decisão;\n",
    "* **`criterion`**: critério de decisão;\n",
    "* **`max_features`**: número máximo de variáveis a serem utilizadas ao procurar melhor divisão;\n",
    "* **`max_depth`**: profundidade máxima da árvore de decisão;\n",
    "* **`min_samples_split`**: número mínimo de amostras necessárias para que uma divisão seja feita;\n",
    "* **`min_samples_leaf`**: número mínimo de amostras necessárias para que uma folha seja formada;\n",
    "* **`bootstrap`**: indica se foi realizada uma reamostragem dos dados na criação dos modelos de árvore de decisão."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_estimators = np.arange(100, 200, step=20)\t\n",
    "criterion = [\"gini\", \"entropy\"]  \n",
    "max_features = [\"auto\", \"log2\"]\n",
    "max_depth = list(np.arange(2, 10, step=2))\n",
    "min_samples_split = np.arange(2, 10, step=2)\n",
    "min_samples_leaf = [2, 4]\n",
    "bootstrap = [True, False]\n",
    "\n",
    "parameters = {\n",
    "    \"n_estimators\": n_estimators,\n",
    "    \"criterion\": criterion,\n",
    "    \"max_features\": max_features,\n",
    "    \"max_depth\": max_depth,\n",
    "    \"min_samples_split\": min_samples_split,\n",
    "    \"min_samples_leaf\": min_samples_leaf,\n",
    "    \"bootstrap\": bootstrap,\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Agora, podemos utilizar o otimizador GridSearchCV para melhorar a mética *Recall*. Para isso, foi criado um objeto GridSearchCV e nele foi passado como parâmetros o nosso modelo, os hipermparametros definidos anteriormente e a métrica a ser otimizada, *Recall*. Bem como, realizamos o ajuste com o método `fit()`.\n",
    "\n",
    "Para saber mais sobre o GridSearchCV e seus parrâmetros você pode consultar a [documentação](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = GridSearchCV(RandomForestClassifier(random_state=SEED), parameters, cv=3, n_jobs=-1, scoring=\"recall\")\n",
    "clf.fit(x_train, y_train)  # ajuste com os dados de treino"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Após o ajuste, conseguimos analisar os resultados obtidos:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "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>mean_fit_time</th>\n",
       "      <th>std_fit_time</th>\n",
       "      <th>mean_score_time</th>\n",
       "      <th>std_score_time</th>\n",
       "      <th>param_bootstrap</th>\n",
       "      <th>param_criterion</th>\n",
       "      <th>param_max_depth</th>\n",
       "      <th>param_max_features</th>\n",
       "      <th>param_min_samples_leaf</th>\n",
       "      <th>param_min_samples_split</th>\n",
       "      <th>param_n_estimators</th>\n",
       "      <th>params</th>\n",
       "      <th>split0_test_score</th>\n",
       "      <th>split1_test_score</th>\n",
       "      <th>split2_test_score</th>\n",
       "      <th>mean_test_score</th>\n",
       "      <th>std_test_score</th>\n",
       "      <th>rank_test_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1230</th>\n",
       "      <td>2.795735</td>\n",
       "      <td>0.710900</td>\n",
       "      <td>0.137522</td>\n",
       "      <td>0.014510</td>\n",
       "      <td>False</td>\n",
       "      <td>entropy</td>\n",
       "      <td>8</td>\n",
       "      <td>auto</td>\n",
       "      <td>4</td>\n",
       "      <td>6</td>\n",
       "      <td>100</td>\n",
       "      <td>{'bootstrap': False, 'criterion': 'entropy', '...</td>\n",
       "      <td>0.894615</td>\n",
       "      <td>0.890085</td>\n",
       "      <td>0.886241</td>\n",
       "      <td>0.890314</td>\n",
       "      <td>0.003423</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>925</th>\n",
       "      <td>1.476667</td>\n",
       "      <td>0.158846</td>\n",
       "      <td>0.115731</td>\n",
       "      <td>0.005462</td>\n",
       "      <td>False</td>\n",
       "      <td>gini</td>\n",
       "      <td>8</td>\n",
       "      <td>log2</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>100</td>\n",
       "      <td>{'bootstrap': False, 'criterion': 'gini', 'max...</td>\n",
       "      <td>0.892308</td>\n",
       "      <td>0.889316</td>\n",
       "      <td>0.889316</td>\n",
       "      <td>0.890313</td>\n",
       "      <td>0.001410</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1245</th>\n",
       "      <td>1.688358</td>\n",
       "      <td>0.039428</td>\n",
       "      <td>0.196186</td>\n",
       "      <td>0.037705</td>\n",
       "      <td>False</td>\n",
       "      <td>entropy</td>\n",
       "      <td>8</td>\n",
       "      <td>log2</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>100</td>\n",
       "      <td>{'bootstrap': False, 'criterion': 'entropy', '...</td>\n",
       "      <td>0.895385</td>\n",
       "      <td>0.886241</td>\n",
       "      <td>0.888547</td>\n",
       "      <td>0.890058</td>\n",
       "      <td>0.003883</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1213</th>\n",
       "      <td>3.567250</td>\n",
       "      <td>0.136539</td>\n",
       "      <td>0.237953</td>\n",
       "      <td>0.032301</td>\n",
       "      <td>False</td>\n",
       "      <td>entropy</td>\n",
       "      <td>8</td>\n",
       "      <td>auto</td>\n",
       "      <td>2</td>\n",
       "      <td>6</td>\n",
       "      <td>160</td>\n",
       "      <td>{'bootstrap': False, 'criterion': 'entropy', '...</td>\n",
       "      <td>0.894615</td>\n",
       "      <td>0.892390</td>\n",
       "      <td>0.883167</td>\n",
       "      <td>0.890058</td>\n",
       "      <td>0.004956</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1217</th>\n",
       "      <td>2.636150</td>\n",
       "      <td>0.202752</td>\n",
       "      <td>0.212012</td>\n",
       "      <td>0.018872</td>\n",
       "      <td>False</td>\n",
       "      <td>entropy</td>\n",
       "      <td>8</td>\n",
       "      <td>auto</td>\n",
       "      <td>2</td>\n",
       "      <td>8</td>\n",
       "      <td>140</td>\n",
       "      <td>{'bootstrap': False, 'criterion': 'entropy', '...</td>\n",
       "      <td>0.898462</td>\n",
       "      <td>0.883935</td>\n",
       "      <td>0.885473</td>\n",
       "      <td>0.889290</td>\n",
       "      <td>0.006516</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      mean_fit_time  std_fit_time  mean_score_time  std_score_time  \\\n",
       "1230       2.795735      0.710900         0.137522        0.014510   \n",
       "925        1.476667      0.158846         0.115731        0.005462   \n",
       "1245       1.688358      0.039428         0.196186        0.037705   \n",
       "1213       3.567250      0.136539         0.237953        0.032301   \n",
       "1217       2.636150      0.202752         0.212012        0.018872   \n",
       "\n",
       "     param_bootstrap param_criterion param_max_depth param_max_features  \\\n",
       "1230           False         entropy               8               auto   \n",
       "925            False            gini               8               log2   \n",
       "1245           False         entropy               8               log2   \n",
       "1213           False         entropy               8               auto   \n",
       "1217           False         entropy               8               auto   \n",
       "\n",
       "     param_min_samples_leaf param_min_samples_split param_n_estimators  \\\n",
       "1230                      4                       6                100   \n",
       "925                       2                       4                100   \n",
       "1245                      2                       4                100   \n",
       "1213                      2                       6                160   \n",
       "1217                      2                       8                140   \n",
       "\n",
       "                                                 params  split0_test_score  \\\n",
       "1230  {'bootstrap': False, 'criterion': 'entropy', '...           0.894615   \n",
       "925   {'bootstrap': False, 'criterion': 'gini', 'max...           0.892308   \n",
       "1245  {'bootstrap': False, 'criterion': 'entropy', '...           0.895385   \n",
       "1213  {'bootstrap': False, 'criterion': 'entropy', '...           0.894615   \n",
       "1217  {'bootstrap': False, 'criterion': 'entropy', '...           0.898462   \n",
       "\n",
       "      split1_test_score  split2_test_score  mean_test_score  std_test_score  \\\n",
       "1230           0.890085           0.886241         0.890314        0.003423   \n",
       "925            0.889316           0.889316         0.890313        0.001410   \n",
       "1245           0.886241           0.888547         0.890058        0.003883   \n",
       "1213           0.892390           0.883167         0.890058        0.004956   \n",
       "1217           0.883935           0.885473         0.889290        0.006516   \n",
       "\n",
       "      rank_test_score  \n",
       "1230                1  \n",
       "925                 2  \n",
       "1245                3  \n",
       "1213                4  \n",
       "1217                5  "
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(clf.cv_results_).sort_values(by=\"rank_test_score\").head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para obtermos os melhores resultados do otimizador utilizamos o comando **`best_params_`**:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'bootstrap': False,\n",
       " 'criterion': 'entropy',\n",
       " 'max_depth': 8,\n",
       " 'max_features': 'auto',\n",
       " 'min_samples_leaf': 4,\n",
       " 'min_samples_split': 6,\n",
       " 'n_estimators': 100}"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Com isso, conseguimos criar um estimador de Random Forest com os melhores hiperparâmetros:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "rforest = RandomForestClassifier(**clf.best_params_, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = rforest.fit(x_train, y_train)  # treinamento com os dados de treino\n",
    "y_pred = rforest.predict(x_test)  # realizando as predições"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Podemos analisar assim, se o modelo realmente obteve resultados melhores na métrica de *Recall* utilizanod a função `plot_matriz_confusao`:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "labels = [\"True Neg\",\"False Pos\",\"False Neg\",\"True Pos\"]\n",
    "categories = [\"Não Churn\", \"Churn\"]\n",
    "plot_matriz_confusao(y_test,\n",
    "                      y_pred,\n",
    "                      group_names=labels,\n",
    "                      categories=categories,\n",
    "                      figsize=(8, 6), \n",
    "                      title=\"Matriz de confusão para o classificador Random Forest\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia de treinamento: 84.89%\n"
     ]
    }
   ],
   "source": [
    "print(f\"Acurácia de treinamento: {rforest.score(x_train, y_train) * 100:.2f}%\")  # Verificando a acurácia de treinamento"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Percebe-se que o recall e as outras métricas aumentaram com o modelo otimizado.\n",
    "\n",
    "<h4>Antes da otimização</h4>\n",
    "\n",
    "|    | Modelo        |   Acurácia Treino |   Acurácia Teste |   Precisão |   Recall |   F1-Score |\n",
    "|---:|:--------------|------------------:|-----------------:|-----------:|---------:|-----------:|\n",
    "|  0 | random forest |          0.809818 |         0.800928 |   0.769395 | 0.849843 |   0.80762  |\n",
    "\n",
    "<h4>Depois da otimização</h4>\n",
    "\n",
    "|    | Modelo        |   Acurácia Treino |   Acurácia Teste |   Precisão |   Recall |   F1-Score |\n",
    "|---:|:--------------|------------------:|-----------------:|-----------:|---------:|-----------:|\n",
    "|  0 | random forest |          0.848859 |         0.822 |   0.793 | 0.864 |   0.827  |\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Desse modo, conseguimos criar um modelo bem otimizado e assim, apresentá-lo ao Alura Voz! :D"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "5920b5758d452ea2be3c84d3b52789bc04ee682949a6b0870d7691434146c726"
  },
  "kernelspec": {
   "display_name": "Python 3.8.13 ('machine_learning')",
   "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": "3.8.13"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}