Viviane Pinto de Souza.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "marine-brain",
   "metadata": {
    "tags": []
   },
   "source": [
    "# Lista de Exercícios - Aula de Redes Neurais e Deep Learning"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "regional-saying",
   "metadata": {},
   "source": [
    "### Questão 1 - O que seria um problema não linearmente separável? Explique utilizando o problema da porta lógica XOR. Argumente sobre o motivo do Perceptron de Rosenblatt (1958) não conseguir resolver este tipo de problema."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "tamil-saturn",
   "metadata": {},
   "source": [
    "Resposta: É aquele problema em que não se pode separar as 2 classes (diferentes, óbvio) no eixo cartesiano bidimensional com apenas uma reta. Um bom exemplo de problema não linearmente separável é a porta lógica XOR (onde a saída é 1 para entradas diferentes). A porta XOR funciona assim:"
   ]
  },
  {
   "attachments": {
    "2e0888bb-447e-4ddd-8363-5dfbbbaa9820.jpg": {
     "image/jpeg": "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"
    }
   },
   "cell_type": "markdown",
   "id": "comfortable-knowing",
   "metadata": {},
   "source": [
    "![xor-11.jpg](attachment:2e0888bb-447e-4ddd-8363-5dfbbbaa9820.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "outstanding-indonesia",
   "metadata": {},
   "source": [
    "Como pode ser visto abaixo, é impossível separar o vermelho (1 - primeira classe) do azul (0 - segunda classe) com apenas uma reta (mesmo tentando de várias formas).\n",
    "\n"
   ]
  },
  {
   "attachments": {
    "37c6fc69-5845-4b57-8a47-e23ca7d49bbf.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "id": "gothic-tracy",
   "metadata": {},
   "source": [
    "![xor setup-1.png](attachment:37c6fc69-5845-4b57-8a47-e23ca7d49bbf.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "comparable-basketball",
   "metadata": {},
   "source": [
    "O Perceptron de Rosenblatt (1958) não irá conseguir resolver problemas não linearmente separáveis (a porta XOR, como tomamos exemplo) porque ele só pode traçar uma reta (ou seja, resolver problemas linearmente separáveis). Através dele não conseguimos resolver o problema da porta XOR, como já dito, mas conseguimos resolver o problema da porta OR e AND. Agora se você montar a porta XOR combinando portas AND e OR, o problema se torna lineramente separável e portanto, conseguimos o Perceptron de Rosenblatt para resolve-lo. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "approximate-solomon",
   "metadata": {},
   "source": [
    "### Questão 2 - Por que uma das estratégias para treinamento da rede neural é baseado no gradiente descendente e qual a relação do gradiente com a atualização dos pesos dos neurônios."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "expanded-drove",
   "metadata": {},
   "source": [
    "Resposta: Através do gradiente é possível simbolizar o erro através de uma função que cresce ou decresce consoante com o erro, isso sendo usado a regra de aprendizagem do erro (que, como já de nosso conhecimento, irá afeiçoar o peso da rede neural). O Backpropagation depende do coeficiente de peso, como também da saída. A derivada parcial da saída vai depender da variação da derivada do peso. Com o resultado em mãos, temos que fazer a minimização do erro. Como o vetor gradiante irá apontar para onde a função cresce, deve-se apontar para o sentido oposto."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dying-pharmacy",
   "metadata": {},
   "source": [
    "### Questão 3 - Considerando o dataset diabetes.csv, o mesmo utilizado nas atividades das aulas de machine learning, faça o treinamento e avalie as métricas de teste, para os modelos Shallow e Deep vistos em aula. Compare os resultados considerando as métricas: Acurácia, Tempo de treino e tempo de teste em relação aos classificadores MLP, KNN e QDA (vistos na aula de machine learning). Crie uma tabela com as métricas e liste os resultados, comente brevemente os resultados apresentados. Considere a Tabela 1 1 para listar os resultados."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "cleared-shelf",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.datasets import make_moons\n",
    "from sklearn.model_selection import train_test_split\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from keras.optimizers import SGD, Adam\n",
    "\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.model_selection import StratifiedShuffleSplit\n",
    "from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "productive-giving",
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_base():\n",
    "    data = pd.read_csv(\"diabetes.csv\")\n",
    "    X = data.iloc[:, :-1].values  \n",
    "    z = data.iloc[:, 8].values\n",
    "    lab = LabelEncoder()\n",
    "    z[:] = lab.fit_transform(z[:])\n",
    "    y = z.astype('int') \n",
    "    \n",
    "    return X,y\n",
    "    \n",
    "    X.shape\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "chemical-roads",
   "metadata": {},
   "outputs": [],
   "source": [
    "X, y = load()\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "seasonal-ebony",
   "metadata": {},
   "outputs": [],
   "source": [
    "acuracia = []\n",
    "tempo_de_treino = []\n",
    "tempo_de_teste = []"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "important-control",
   "metadata": {},
   "source": [
    "# *Modelo Shallow*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "danish-nashville",
   "metadata": {},
   "outputs": [],
   "source": [
    "modelo_shallow = Sequential()\n",
    "modelo_shallow.add(Dense(1, input_shape=(8,), activation='sigmoid'))\n",
    "modelo_shallow.compile(Adam(lr=0.05), 'binary_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "frozen-circular",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "17/17 [==============================] - 1s 2ms/step - loss: 32.7743 - accuracy: 0.4430\n",
      "Epoch 2/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 8.7669 - accuracy: 0.6702\n",
      "Epoch 3/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 4.0670 - accuracy: 0.6659\n",
      "Epoch 4/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 3.5100 - accuracy: 0.6335\n",
      "Epoch 5/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 2.8014 - accuracy: 0.6225\n",
      "Epoch 6/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.7881 - accuracy: 0.7018\n",
      "Epoch 7/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.7866 - accuracy: 0.5986\n",
      "Epoch 8/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.0803 - accuracy: 0.6405\n",
      "Epoch 9/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.0662 - accuracy: 0.5686\n",
      "Epoch 10/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9546 - accuracy: 0.6206\n",
      "Epoch 11/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9048 - accuracy: 0.6400\n",
      "Epoch 12/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6811 - accuracy: 0.6710\n",
      "Epoch 13/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.8735 - accuracy: 0.6123\n",
      "Epoch 14/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9351 - accuracy: 0.6234\n",
      "Epoch 15/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7971 - accuracy: 0.5981\n",
      "Epoch 16/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6955 - accuracy: 0.6537\n",
      "Epoch 17/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8567 - accuracy: 0.5784\n",
      "Epoch 18/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7850 - accuracy: 0.6637\n",
      "Epoch 19/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8395 - accuracy: 0.6162\n",
      "Epoch 20/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7867 - accuracy: 0.6272\n",
      "Epoch 21/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6098 - accuracy: 0.7139\n",
      "Epoch 22/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6892 - accuracy: 0.6641\n",
      "Epoch 23/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.1487 - accuracy: 0.5846\n",
      "Epoch 24/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.4214 - accuracy: 0.5936\n",
      "Epoch 25/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.0847 - accuracy: 0.6306\n",
      "Epoch 26/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.1103 - accuracy: 0.6203\n",
      "Epoch 27/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8855 - accuracy: 0.5974\n",
      "Epoch 28/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.0090 - accuracy: 0.6643\n",
      "Epoch 29/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.7812 - accuracy: 0.6511\n",
      "Epoch 30/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.1803 - accuracy: 0.5726\n",
      "Epoch 31/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7280 - accuracy: 0.6815\n",
      "Epoch 32/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6393 - accuracy: 0.6963\n",
      "Epoch 33/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6824 - accuracy: 0.6589\n",
      "Epoch 34/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8918 - accuracy: 0.6634\n",
      "Epoch 35/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9897 - accuracy: 0.5913\n",
      "Epoch 36/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5681 - accuracy: 0.7377\n",
      "Epoch 37/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9307 - accuracy: 0.6049\n",
      "Epoch 38/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7574 - accuracy: 0.6912\n",
      "Epoch 39/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6873 - accuracy: 0.7063\n",
      "Epoch 40/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.7501 - accuracy: 0.6410\n",
      "Epoch 41/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6840 - accuracy: 0.6517\n",
      "Epoch 42/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.7532 - accuracy: 0.6455\n",
      "Epoch 43/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.7719 - accuracy: 0.6656\n",
      "Epoch 44/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9537 - accuracy: 0.6369\n",
      "Epoch 45/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7604 - accuracy: 0.6886\n",
      "Epoch 46/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.7404 - accuracy: 0.6964\n",
      "Epoch 47/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8486 - accuracy: 0.6464\n",
      "Epoch 48/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5697 - accuracy: 0.7195\n",
      "Epoch 49/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6225 - accuracy: 0.6907\n",
      "Epoch 50/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8659 - accuracy: 0.6586\n",
      "Epoch 51/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9610 - accuracy: 0.6896\n",
      "Epoch 52/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6770 - accuracy: 0.6875\n",
      "Epoch 53/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8641 - accuracy: 0.6246\n",
      "Epoch 54/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9974 - accuracy: 0.6715\n",
      "Epoch 55/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8433 - accuracy: 0.6260\n",
      "Epoch 56/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5819 - accuracy: 0.7085\n",
      "Epoch 57/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7791 - accuracy: 0.6299\n",
      "Epoch 58/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6149 - accuracy: 0.7363\n",
      "Epoch 59/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.7743 - accuracy: 0.6509\n",
      "Epoch 60/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.5416 - accuracy: 0.7389\n",
      "Epoch 61/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8301 - accuracy: 0.6578\n",
      "Epoch 62/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7616 - accuracy: 0.6809\n",
      "Epoch 63/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8382 - accuracy: 0.6278\n",
      "Epoch 64/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6952 - accuracy: 0.7098\n",
      "Epoch 65/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5717 - accuracy: 0.7536\n",
      "Epoch 66/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.8059 - accuracy: 0.6667\n",
      "Epoch 67/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6148 - accuracy: 0.6953\n",
      "Epoch 68/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9558 - accuracy: 0.6875\n",
      "Epoch 69/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.4639 - accuracy: 0.5702\n",
      "Epoch 70/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.2616 - accuracy: 0.6142\n",
      "Epoch 71/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8279 - accuracy: 0.6604\n",
      "Epoch 72/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.1740 - accuracy: 0.5948\n",
      "Epoch 73/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5988 - accuracy: 0.7350\n",
      "Epoch 74/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7245 - accuracy: 0.6838\n",
      "Epoch 75/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5513 - accuracy: 0.7619\n",
      "Epoch 76/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5591 - accuracy: 0.7302\n",
      "Epoch 77/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6467 - accuracy: 0.6967\n",
      "Epoch 78/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9086 - accuracy: 0.6409\n",
      "Epoch 79/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6398 - accuracy: 0.7154\n",
      "Epoch 80/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6686 - accuracy: 0.6929\n",
      "Epoch 81/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6722 - accuracy: 0.7089\n",
      "Epoch 82/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.7035 - accuracy: 0.6724\n",
      "Epoch 83/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.8095 - accuracy: 0.6744\n",
      "Epoch 84/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 1.2811 - accuracy: 0.6369\n",
      "Epoch 85/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 1.3822 - accuracy: 0.5998\n",
      "Epoch 86/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8074 - accuracy: 0.6696\n",
      "Epoch 87/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6507 - accuracy: 0.7461\n",
      "Epoch 88/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7096 - accuracy: 0.7045\n",
      "Epoch 89/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7166 - accuracy: 0.6460\n",
      "Epoch 90/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7620 - accuracy: 0.6919\n",
      "Epoch 91/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9958 - accuracy: 0.6277\n",
      "Epoch 92/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5800 - accuracy: 0.7505\n",
      "Epoch 93/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.7020 - accuracy: 0.6854\n",
      "Epoch 94/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7228 - accuracy: 0.6999\n",
      "Epoch 95/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8414 - accuracy: 0.6540\n",
      "Epoch 96/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6454 - accuracy: 0.7151\n",
      "Epoch 97/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6139 - accuracy: 0.7221\n",
      "Epoch 98/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8038 - accuracy: 0.7072\n",
      "Epoch 99/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.8398 - accuracy: 0.6669\n",
      "Epoch 100/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8071 - accuracy: 0.7007\n",
      "Epoch 101/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.5783 - accuracy: 0.7447\n",
      "Epoch 102/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5769 - accuracy: 0.7320\n",
      "Epoch 103/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5467 - accuracy: 0.7314\n",
      "Epoch 104/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5394 - accuracy: 0.7638\n",
      "Epoch 105/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7366 - accuracy: 0.6863\n",
      "Epoch 106/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6544 - accuracy: 0.7114\n",
      "Epoch 107/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6918 - accuracy: 0.6910\n",
      "Epoch 108/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.4902 - accuracy: 0.7666\n",
      "Epoch 109/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5906 - accuracy: 0.7298\n",
      "Epoch 110/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5636 - accuracy: 0.7565\n",
      "Epoch 111/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7343 - accuracy: 0.6920\n",
      "Epoch 112/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5745 - accuracy: 0.7484\n",
      "Epoch 113/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7424 - accuracy: 0.6880\n",
      "Epoch 114/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7930 - accuracy: 0.6906\n",
      "Epoch 115/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7212 - accuracy: 0.6666\n",
      "Epoch 116/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8241 - accuracy: 0.6869\n",
      "Epoch 117/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6194 - accuracy: 0.7088\n",
      "Epoch 118/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.2631 - accuracy: 0.6433\n",
      "Epoch 119/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.3865 - accuracy: 0.6626\n",
      "Epoch 120/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9117 - accuracy: 0.6347\n",
      "Epoch 121/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.2347 - accuracy: 0.6614\n",
      "Epoch 122/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.0556 - accuracy: 0.6521\n",
      "Epoch 123/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6898 - accuracy: 0.6949\n",
      "Epoch 124/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.2721 - accuracy: 0.6512\n",
      "Epoch 125/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8302 - accuracy: 0.6977\n",
      "Epoch 126/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.3028 - accuracy: 0.5720\n",
      "Epoch 127/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6654 - accuracy: 0.7246\n",
      "Epoch 128/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6183 - accuracy: 0.7400\n",
      "Epoch 129/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.0109 - accuracy: 0.6497\n",
      "Epoch 130/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7700 - accuracy: 0.7177\n",
      "Epoch 131/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7990 - accuracy: 0.7108\n",
      "Epoch 132/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6677 - accuracy: 0.7395\n",
      "Epoch 133/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5340 - accuracy: 0.7331\n",
      "Epoch 134/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5081 - accuracy: 0.7708\n",
      "Epoch 135/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7410 - accuracy: 0.7185\n",
      "Epoch 136/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5762 - accuracy: 0.7567\n",
      "Epoch 137/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.0103 - accuracy: 0.6813\n",
      "Epoch 138/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6822 - accuracy: 0.7186\n",
      "Epoch 139/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5052 - accuracy: 0.7709\n",
      "Epoch 140/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6978 - accuracy: 0.7169\n",
      "Epoch 141/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6340 - accuracy: 0.7164\n",
      "Epoch 142/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8442 - accuracy: 0.7027\n",
      "Epoch 143/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.0264 - accuracy: 0.6922\n",
      "Epoch 144/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7267 - accuracy: 0.6818\n",
      "Epoch 145/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9026 - accuracy: 0.7319\n",
      "Epoch 146/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5095 - accuracy: 0.7613\n",
      "Epoch 147/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6778 - accuracy: 0.6967\n",
      "Epoch 148/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6448 - accuracy: 0.7266\n",
      "Epoch 149/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7083 - accuracy: 0.7252\n",
      "Epoch 150/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5783 - accuracy: 0.7323\n",
      "Epoch 151/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7624 - accuracy: 0.7194\n",
      "Epoch 152/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6652 - accuracy: 0.7028\n",
      "Epoch 153/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7345 - accuracy: 0.6937\n",
      "Epoch 154/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6720 - accuracy: 0.7264\n",
      "Epoch 155/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9871 - accuracy: 0.6620\n",
      "Epoch 156/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7337 - accuracy: 0.7096\n",
      "Epoch 157/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5350 - accuracy: 0.7367\n",
      "Epoch 158/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7779 - accuracy: 0.7087\n",
      "Epoch 159/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8063 - accuracy: 0.6886\n",
      "Epoch 160/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5829 - accuracy: 0.7408\n",
      "Epoch 161/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5321 - accuracy: 0.7391\n",
      "Epoch 162/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.5887 - accuracy: 0.7308\n",
      "Epoch 163/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.5025 - accuracy: 0.7753\n",
      "Epoch 164/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6353 - accuracy: 0.7261\n",
      "Epoch 165/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6438 - accuracy: 0.7432\n",
      "Epoch 166/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5331 - accuracy: 0.7447\n",
      "Epoch 167/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8230 - accuracy: 0.6968\n",
      "Epoch 168/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6583 - accuracy: 0.7166\n",
      "Epoch 169/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6332 - accuracy: 0.7296\n",
      "Epoch 170/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6835 - accuracy: 0.7006\n",
      "Epoch 171/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6969 - accuracy: 0.7587\n",
      "Epoch 172/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5750 - accuracy: 0.7747\n",
      "Epoch 173/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6535 - accuracy: 0.7577\n",
      "Epoch 174/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6270 - accuracy: 0.7228\n",
      "Epoch 175/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5751 - accuracy: 0.7518\n",
      "Epoch 176/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9351 - accuracy: 0.6676\n",
      "Epoch 177/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8505 - accuracy: 0.7093\n",
      "Epoch 178/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7212 - accuracy: 0.7070\n",
      "Epoch 179/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 1.1803 - accuracy: 0.6679\n",
      "Epoch 180/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6727 - accuracy: 0.7406\n",
      "Epoch 181/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5454 - accuracy: 0.7441\n",
      "Epoch 182/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6996 - accuracy: 0.7086\n",
      "Epoch 183/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6329 - accuracy: 0.7128\n",
      "Epoch 184/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5200 - accuracy: 0.7820\n",
      "Epoch 185/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5227 - accuracy: 0.7720\n",
      "Epoch 186/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6348 - accuracy: 0.7236\n",
      "Epoch 187/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.9289 - accuracy: 0.6732\n",
      "Epoch 188/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5698 - accuracy: 0.7627\n",
      "Epoch 189/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5753 - accuracy: 0.7392\n",
      "Epoch 190/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.5873 - accuracy: 0.7275\n",
      "Epoch 191/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6772 - accuracy: 0.7187\n",
      "Epoch 192/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6211 - accuracy: 0.7172\n",
      "Epoch 193/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7031 - accuracy: 0.7015\n",
      "Epoch 194/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8029 - accuracy: 0.6948\n",
      "Epoch 195/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.8997 - accuracy: 0.6775\n",
      "Epoch 196/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7023 - accuracy: 0.7171\n",
      "Epoch 197/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7422 - accuracy: 0.7214\n",
      "Epoch 198/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7873 - accuracy: 0.7141\n",
      "Epoch 199/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.7520 - accuracy: 0.6792\n",
      "Epoch 200/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.8243 - accuracy: 0.7083\n"
     ]
    }
   ],
   "source": [
    "ti_treino = time.time()\n",
    "modelo_shallow.fit(X_train, y_train, epochs=200, verbose=1)\n",
    "tf_treino = time.time()\n",
    "\n",
    "time_treino_shallow = tf_treino - ti_treino\n",
    "tempo_de_treino.append(time_treino_shallow)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "humanitarian-dealing",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8/8 [==============================] - 0s 2ms/step - loss: 0.7263 - accuracy: 0.6970\n"
     ]
    }
   ],
   "source": [
    "ti_teste = time.time()\n",
    "resultado_shallow = modelo_shallow.evaluate(X_test, y_test)\n",
    "tf_teste = time.time()\n",
    "\n",
    "time_teste_shallow = tf_teste - ti_teste\n",
    "\n",
    "tempo_de_teste.append(time_teste_shallow)\n",
    "acuracia.append(resultado_shallow[1])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "commercial-silver",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia:  0.6969696879386902\n",
      "Tempo de treino:  9.996330976486206\n",
      "Tempo de Teste:  0.09597444534301758\n"
     ]
    }
   ],
   "source": [
    "print('Acurácia: ', resultado_shallow[1])\n",
    "print('Tempo de treino: ', time_treino_shallow)\n",
    "print('Tempo de Teste: ', time_teste_shallow)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "convenient-bible",
   "metadata": {},
   "source": [
    "# *Model Deep*\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "synthetic-allen",
   "metadata": {},
   "outputs": [],
   "source": [
    "modelo_deep = Sequential()\n",
    "modelo_deep.add(Dense(4, input_shape=(8,), activation='tanh')) \n",
    "modelo_deep.add(Dense(2, input_shape=(16,), activation='tanh')) \n",
    "modelo_deep.add(Dense(1, input_shape=(8,), activation='sigmoid')) \n",
    "modelo_deep.compile(optimizer=Adam(lr=0.05), loss='binary_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "animated-advocate",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "17/17 [==============================] - 1s 2ms/step - loss: 0.6696 - accuracy: 0.6547\n",
      "Epoch 2/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6458 - accuracy: 0.6550\n",
      "Epoch 3/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6394 - accuracy: 0.6655\n",
      "Epoch 4/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6467 - accuracy: 0.6548\n",
      "Epoch 5/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6621 - accuracy: 0.6302\n",
      "Epoch 6/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6669 - accuracy: 0.6250\n",
      "Epoch 7/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6540 - accuracy: 0.6413\n",
      "Epoch 8/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6521 - accuracy: 0.6467\n",
      "Epoch 9/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6494 - accuracy: 0.6498\n",
      "Epoch 10/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6625 - accuracy: 0.6269\n",
      "Epoch 11/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6502 - accuracy: 0.6472\n",
      "Epoch 12/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6486 - accuracy: 0.6497\n",
      "Epoch 13/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6439 - accuracy: 0.6577\n",
      "Epoch 14/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6708 - accuracy: 0.6235\n",
      "Epoch 15/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6696 - accuracy: 0.6254\n",
      "Epoch 16/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6528 - accuracy: 0.6463\n",
      "Epoch 17/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6469 - accuracy: 0.6530\n",
      "Epoch 18/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6484 - accuracy: 0.6492\n",
      "Epoch 19/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6664 - accuracy: 0.6205\n",
      "Epoch 20/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6399 - accuracy: 0.6671\n",
      "Epoch 21/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6627 - accuracy: 0.6306\n",
      "Epoch 22/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6356 - accuracy: 0.6705\n",
      "Epoch 23/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6583 - accuracy: 0.6377\n",
      "Epoch 24/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6606 - accuracy: 0.6300\n",
      "Epoch 25/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6311 - accuracy: 0.6773\n",
      "Epoch 26/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6462 - accuracy: 0.6566\n",
      "Epoch 27/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6420 - accuracy: 0.6603\n",
      "Epoch 28/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6604 - accuracy: 0.6320\n",
      "Epoch 29/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6465 - accuracy: 0.6573\n",
      "Epoch 30/200\n",
      "17/17 [==============================] - 0s 4ms/step - loss: 0.6453 - accuracy: 0.6544\n",
      "Epoch 31/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6601 - accuracy: 0.6292\n",
      "Epoch 32/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6625 - accuracy: 0.6305\n",
      "Epoch 33/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6638 - accuracy: 0.6246\n",
      "Epoch 34/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6291 - accuracy: 0.6792\n",
      "Epoch 35/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6487 - accuracy: 0.6488\n",
      "Epoch 36/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6548 - accuracy: 0.6379\n",
      "Epoch 37/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6668 - accuracy: 0.6240\n",
      "Epoch 38/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6488 - accuracy: 0.6494\n",
      "Epoch 39/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6649 - accuracy: 0.6245\n",
      "Epoch 40/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6478 - accuracy: 0.6501\n",
      "Epoch 41/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6409 - accuracy: 0.6614\n",
      "Epoch 42/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6428 - accuracy: 0.6584\n",
      "Epoch 43/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6359 - accuracy: 0.6709\n",
      "Epoch 44/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6515 - accuracy: 0.6446\n",
      "Epoch 45/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6441 - accuracy: 0.6565\n",
      "Epoch 46/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6436 - accuracy: 0.6573\n",
      "Epoch 47/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6293 - accuracy: 0.6824\n",
      "Epoch 48/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6353 - accuracy: 0.6715\n",
      "Epoch 49/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6509 - accuracy: 0.6458\n",
      "Epoch 50/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6316 - accuracy: 0.6760\n",
      "Epoch 51/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6400 - accuracy: 0.6646\n",
      "Epoch 52/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6614 - accuracy: 0.6275\n",
      "Epoch 53/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6677 - accuracy: 0.6191\n",
      "Epoch 54/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6597 - accuracy: 0.6349\n",
      "Epoch 55/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6272 - accuracy: 0.6805\n",
      "Epoch 56/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6495 - accuracy: 0.6490\n",
      "Epoch 57/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6481 - accuracy: 0.6506\n",
      "Epoch 58/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6432 - accuracy: 0.6582\n",
      "Epoch 59/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6433 - accuracy: 0.6573\n",
      "Epoch 60/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6564 - accuracy: 0.6362\n",
      "Epoch 61/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6660 - accuracy: 0.6186\n",
      "Epoch 62/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6576 - accuracy: 0.6356\n",
      "Epoch 63/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6343 - accuracy: 0.6711\n",
      "Epoch 64/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6714 - accuracy: 0.6131\n",
      "Epoch 65/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6693 - accuracy: 0.6217\n",
      "Epoch 66/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6534 - accuracy: 0.6420\n",
      "Epoch 67/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6520 - accuracy: 0.6439\n",
      "Epoch 68/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6359 - accuracy: 0.6705\n",
      "Epoch 69/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6467 - accuracy: 0.6534\n",
      "Epoch 70/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6510 - accuracy: 0.6446\n",
      "Epoch 71/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6620 - accuracy: 0.6287\n",
      "Epoch 72/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6521 - accuracy: 0.6433\n",
      "Epoch 73/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6553 - accuracy: 0.6392\n",
      "Epoch 74/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6504 - accuracy: 0.6469\n",
      "Epoch 75/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6476 - accuracy: 0.6509\n",
      "Epoch 76/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6514 - accuracy: 0.6447\n",
      "Epoch 77/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6597 - accuracy: 0.6324\n",
      "Epoch 78/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6523 - accuracy: 0.6430\n",
      "Epoch 79/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6483 - accuracy: 0.6493\n",
      "Epoch 80/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6379 - accuracy: 0.6657\n",
      "Epoch 81/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6455 - accuracy: 0.6542\n",
      "Epoch 82/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6617 - accuracy: 0.6274\n",
      "Epoch 83/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6427 - accuracy: 0.6588\n",
      "Epoch 84/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6506 - accuracy: 0.6473\n",
      "Epoch 85/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6611 - accuracy: 0.6278\n",
      "Epoch 86/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6508 - accuracy: 0.6446\n",
      "Epoch 87/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6462 - accuracy: 0.6541\n",
      "Epoch 88/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6688 - accuracy: 0.6164\n",
      "Epoch 89/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6510 - accuracy: 0.6472\n",
      "Epoch 90/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6564 - accuracy: 0.6367\n",
      "Epoch 91/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6377 - accuracy: 0.6680\n",
      "Epoch 92/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6380 - accuracy: 0.6682\n",
      "Epoch 93/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6583 - accuracy: 0.6322\n",
      "Epoch 94/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6506 - accuracy: 0.6476\n",
      "Epoch 95/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6568 - accuracy: 0.6360\n",
      "Epoch 96/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6400 - accuracy: 0.6632\n",
      "Epoch 97/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6337 - accuracy: 0.6731\n",
      "Epoch 98/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6462 - accuracy: 0.6537\n",
      "Epoch 99/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6537 - accuracy: 0.6398\n",
      "Epoch 100/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6463 - accuracy: 0.6526\n",
      "Epoch 101/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6479 - accuracy: 0.6500\n",
      "Epoch 102/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6390 - accuracy: 0.6664\n",
      "Epoch 103/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6513 - accuracy: 0.6451\n",
      "Epoch 104/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6539 - accuracy: 0.6414\n",
      "Epoch 105/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6558 - accuracy: 0.6388\n",
      "Epoch 106/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6453 - accuracy: 0.6547\n",
      "Epoch 107/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6395 - accuracy: 0.6640\n",
      "Epoch 108/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6448 - accuracy: 0.6561\n",
      "Epoch 109/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6567 - accuracy: 0.6357\n",
      "Epoch 110/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6522 - accuracy: 0.6417\n",
      "Epoch 111/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6595 - accuracy: 0.6378\n",
      "Epoch 112/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6464 - accuracy: 0.6529\n",
      "Epoch 113/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6489 - accuracy: 0.6480\n",
      "Epoch 114/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6242 - accuracy: 0.6869\n",
      "Epoch 115/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6476 - accuracy: 0.6520\n",
      "Epoch 116/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6461 - accuracy: 0.6553\n",
      "Epoch 117/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6674 - accuracy: 0.6166\n",
      "Epoch 118/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6396 - accuracy: 0.6654\n",
      "Epoch 119/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6415 - accuracy: 0.6607\n",
      "Epoch 120/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6490 - accuracy: 0.6496\n",
      "Epoch 121/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6522 - accuracy: 0.6435\n",
      "Epoch 122/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6570 - accuracy: 0.6377\n",
      "Epoch 123/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6353 - accuracy: 0.6701\n",
      "Epoch 124/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6779 - accuracy: 0.6086\n",
      "Epoch 125/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6306 - accuracy: 0.6797\n",
      "Epoch 126/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6675 - accuracy: 0.6213\n",
      "Epoch 127/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6567 - accuracy: 0.6351\n",
      "Epoch 128/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6368 - accuracy: 0.6650\n",
      "Epoch 129/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6615 - accuracy: 0.6266\n",
      "Epoch 130/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6654 - accuracy: 0.6227\n",
      "Epoch 131/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6395 - accuracy: 0.6629\n",
      "Epoch 132/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6669 - accuracy: 0.6224\n",
      "Epoch 133/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6636 - accuracy: 0.6254\n",
      "Epoch 134/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6347 - accuracy: 0.6701\n",
      "Epoch 135/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6687 - accuracy: 0.6183\n",
      "Epoch 136/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6523 - accuracy: 0.6444\n",
      "Epoch 137/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6621 - accuracy: 0.6310\n",
      "Epoch 138/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6326 - accuracy: 0.6789\n",
      "Epoch 139/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6444 - accuracy: 0.6571\n",
      "Epoch 140/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6621 - accuracy: 0.6265\n",
      "Epoch 141/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6494 - accuracy: 0.6490\n",
      "Epoch 142/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6501 - accuracy: 0.6468\n",
      "Epoch 143/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6595 - accuracy: 0.6318\n",
      "Epoch 144/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6545 - accuracy: 0.6419\n",
      "Epoch 145/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6630 - accuracy: 0.6272\n",
      "Epoch 146/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6621 - accuracy: 0.6333\n",
      "Epoch 147/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6552 - accuracy: 0.6407\n",
      "Epoch 148/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6655 - accuracy: 0.6242\n",
      "Epoch 149/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6607 - accuracy: 0.6325\n",
      "Epoch 150/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6463 - accuracy: 0.6536\n",
      "Epoch 151/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6550 - accuracy: 0.6381\n",
      "Epoch 152/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6403 - accuracy: 0.6625\n",
      "Epoch 153/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6571 - accuracy: 0.6355\n",
      "Epoch 154/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6436 - accuracy: 0.6577\n",
      "Epoch 155/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6410 - accuracy: 0.6629\n",
      "Epoch 156/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6471 - accuracy: 0.6519\n",
      "Epoch 157/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6518 - accuracy: 0.6436\n",
      "Epoch 158/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6549 - accuracy: 0.6428\n",
      "Epoch 159/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6510 - accuracy: 0.6454\n",
      "Epoch 160/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6704 - accuracy: 0.6325\n",
      "Epoch 161/200\n",
      "17/17 [==============================] - ETA: 0s - loss: 0.7374 - accuracy: 0.46 - 0s 2ms/step - loss: 0.6583 - accuracy: 0.6298\n",
      "Epoch 162/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6470 - accuracy: 0.6554\n",
      "Epoch 163/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6546 - accuracy: 0.6485\n",
      "Epoch 164/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6484 - accuracy: 0.6498\n",
      "Epoch 165/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6408 - accuracy: 0.6641\n",
      "Epoch 166/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6452 - accuracy: 0.6570\n",
      "Epoch 167/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6448 - accuracy: 0.6551\n",
      "Epoch 168/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6582 - accuracy: 0.6331\n",
      "Epoch 169/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6432 - accuracy: 0.6573\n",
      "Epoch 170/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6682 - accuracy: 0.6158\n",
      "Epoch 171/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6602 - accuracy: 0.6341\n",
      "Epoch 172/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6491 - accuracy: 0.6501\n",
      "Epoch 173/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6569 - accuracy: 0.6374\n",
      "Epoch 174/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6609 - accuracy: 0.6355\n",
      "Epoch 175/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6251 - accuracy: 0.6967\n",
      "Epoch 176/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6629 - accuracy: 0.6304\n",
      "Epoch 177/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6299 - accuracy: 0.6804\n",
      "Epoch 178/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6542 - accuracy: 0.6587\n",
      "Epoch 179/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6478 - accuracy: 0.6524\n",
      "Epoch 180/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6681 - accuracy: 0.6275\n",
      "Epoch 181/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6459 - accuracy: 0.6609\n",
      "Epoch 182/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6438 - accuracy: 0.6579\n",
      "Epoch 183/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6449 - accuracy: 0.6572\n",
      "Epoch 184/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6363 - accuracy: 0.6684\n",
      "Epoch 185/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6487 - accuracy: 0.6505\n",
      "Epoch 186/200\n",
      "17/17 [==============================] - 0s 3ms/step - loss: 0.6466 - accuracy: 0.6527\n",
      "Epoch 187/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6525 - accuracy: 0.6426\n",
      "Epoch 188/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6582 - accuracy: 0.6324\n",
      "Epoch 189/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6229 - accuracy: 0.6917\n",
      "Epoch 190/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6617 - accuracy: 0.6312\n",
      "Epoch 191/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6457 - accuracy: 0.6570\n",
      "Epoch 192/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6386 - accuracy: 0.6671\n",
      "Epoch 193/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6491 - accuracy: 0.6529\n",
      "Epoch 194/200\n",
      "17/17 [==============================] - 0s 1ms/step - loss: 0.6503 - accuracy: 0.6457\n",
      "Epoch 195/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6467 - accuracy: 0.6524\n",
      "Epoch 196/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6732 - accuracy: 0.6137\n",
      "Epoch 197/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6682 - accuracy: 0.6187\n",
      "Epoch 198/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6423 - accuracy: 0.6613\n",
      "Epoch 199/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6290 - accuracy: 0.6835\n",
      "Epoch 200/200\n",
      "17/17 [==============================] - 0s 2ms/step - loss: 0.6210 - accuracy: 0.6948\n"
     ]
    }
   ],
   "source": [
    "ti_treino = time.time()\n",
    "modelo_deep.fit(X_train, y_train, epochs=200, verbose=1)\n",
    "tf_treino = time.time()\n",
    "\n",
    "time_treino_deep = tf_treino - ti_treino\n",
    "\n",
    "tempo_de_treino.append(time_treino_deep)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "peripheral-century",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8/8 [==============================] - 0s 2ms/step - loss: 0.6471 - accuracy: 0.6537\n"
     ]
    }
   ],
   "source": [
    "ti_teste = time.time()\n",
    "resultado_deep = modelo_deep.evaluate(X_test, y_test)\n",
    "tf_teste = time.time()\n",
    "\n",
    "time_teste_deep = tf_teste - ti_teste\n",
    "\n",
    "teste.append(time_teste_deep)\n",
    "acuracia.append(resultado_deep[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "thermal-netscape",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia:  0.6536796689033508\n",
      "Tempo de Treino:  8.715456485748291\n",
      "Tempo de Teste:  0.21568632125854492\n"
     ]
    }
   ],
   "source": [
    "print('Acurácia: ', resultado_deep[1])\n",
    "print('Tempo de Treino: ', time_treino_deep)\n",
    "print('Tempo de Teste: ', time_teste_deep)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "upper-admission",
   "metadata": {},
   "source": [
    "# *Classificador MLP*\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "preceding-beads",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "*** MLP ***\n",
      "\n",
      "Acurácia MLP:  0.7575757575757576\n",
      "Tempo de Treino:  2.3361668586730957\n",
      "Tempo de Teste:  0.0013430118560791016\n"
     ]
    }
   ],
   "source": [
    "mlp = MLPClassifier(solver='adam', learning_rate='adaptive', hidden_layer_sizes = (100,), \n",
    "                        max_iter=9000, learning_rate_init=5e-04, tol=1e-4)\n",
    "\n",
    "tit_mlp = time.time()\n",
    "mlp.fit(X, y)\n",
    "tft_mlp = time.time()\n",
    "    \n",
    "ti_mlp = time.time()\n",
    "mlp_predicted = mlp.predict(X_test)\n",
    "tf_mlp = time.time()\n",
    "    \n",
    "    \n",
    "print('*** MLP ***\\n')\n",
    "acu_mlp = accuracy_score(y_test, mlp_predicted)\n",
    "acuracia.append(acu_mlp)\n",
    "print('Acurácia MLP: ', acu_mlp)\n",
    "tempo_de_treino_mlp = tft_mlp - tit_mlp\n",
    "tempo_de_treino.append(tempo_de_treino_mlp)\n",
    "print('Tempo de Treino: ', tempo_de_treino_mlp)\n",
    "tempo_de_teste_mlp = tf_mlp - ti_mlp\n",
    "tempo_de_teste.append(tempo_de_teste_mlp)\n",
    "print(\"Tempo de Teste: \", tempo_de_teste_mlp)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "structural-macro",
   "metadata": {},
   "source": [
    "# *Classificador KNN*\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "advance-noise",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "*** KNN ***\n",
      "\n",
      "Acurácia MLP:  0.7575757575757576\n",
      "Tempo de Treino:  0.06163287162780762\n",
      "Tempo de Teste:  0.02576279640197754\n"
     ]
    }
   ],
   "source": [
    "neigh = KNeighborsClassifier(n_neighbors=3)\n",
    "\n",
    "tit_knn = time.time()\n",
    "neigh.fit(X_train, y_train)\n",
    "tft_knn = time.time()\n",
    "\n",
    "ti_knn = time.time()\n",
    "knn_predicted = neigh.predict(X_test)\n",
    "tf_knn = time.time()\n",
    "\n",
    "print('*** KNN ***\\n')\n",
    "acu_knn = accuracy_score(y_test, knn_predicted)\n",
    "acuracia.append(acu_knn)\n",
    "print('Acurácia MLP: ', acu_mlp)\n",
    "tempo_de_treino_knn = tft_knn - tit_knn\n",
    "tempo_de_treino.append(tempo_de_treino_knn)\n",
    "print('Tempo de Treino: ', tempo_de_treino_knn)\n",
    "tempo_de_teste_knn = tf_knn - ti_knn\n",
    "tempo_de_teste.append(tempo_de_teste_knn)\n",
    "print(\"Tempo de Teste: \", tempo_de_teste_knn)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "forced-spectacular",
   "metadata": {},
   "source": [
    "# *Classificador QDA*\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "negative-disposition",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "*** QDA ***\n",
      "\n",
      "Acurácia QDA : 0.7792207792207793\n",
      "Tempo de Treino:  0.002786397933959961\n",
      "Tempo de Teste:  0.0007796287536621094\n"
     ]
    }
   ],
   "source": [
    "qda = QuadraticDiscriminantAnalysis(store_covariance=True)\n",
    "tit_qda = time.time()\n",
    "qda.fit(X, y)\n",
    "tft_qda = time.time()\n",
    "    \n",
    "ti_qda = time.time()\n",
    "qda_predicted = qda.predict(X_test)\n",
    "tf_qda = time.time()\n",
    "    \n",
    "    \n",
    "print('*** QDA ***\\n')\n",
    "acu_qda = accuracy_score(y_test, qda_predicted)\n",
    "acuracia.append(acu_qda)\n",
    "print('Acurácia QDA :', acu_qda)\n",
    "tempo_de_treino_qda = tft_qda - tit_qda\n",
    "tempo_de_treino.append(tempo_de_treino_qda)\n",
    "print('Tempo de Treino: ', tempo_de_treino_qda)\n",
    "tempo_de_teste_qda = tf_qda - ti_qda\n",
    "tempo_de_teste.append(tempo_de_teste_qda)\n",
    "print('Tempo de Teste: ', tempo_de_teste_qda)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fitting-roberts",
   "metadata": {},
   "source": [
    "# *Tabela com as Métricas (listar os resultados)*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "suspected-hardware",
   "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>Classificador</th>\n",
       "      <th>Acurácia</th>\n",
       "      <th>Tempo de Treino</th>\n",
       "      <th>Tempo de Teste</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Modelo Shallow</td>\n",
       "      <td>0.696970</td>\n",
       "      <td>9.996331</td>\n",
       "      <td>0.095727</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Modelo Deep</td>\n",
       "      <td>0.696970</td>\n",
       "      <td>8.715456</td>\n",
       "      <td>0.095974</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>MLP</td>\n",
       "      <td>0.696970</td>\n",
       "      <td>0.002369</td>\n",
       "      <td>0.000647</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>KNN</td>\n",
       "      <td>0.653680</td>\n",
       "      <td>2.420266</td>\n",
       "      <td>0.000878</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>QDA</td>\n",
       "      <td>0.779221</td>\n",
       "      <td>1.620014</td>\n",
       "      <td>0.000993</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Classificador  Acurácia  Tempo de Treino  Tempo de Teste\n",
       "0  Modelo Shallow  0.696970         9.996331        0.095727\n",
       "1     Modelo Deep  0.696970         8.715456        0.095974\n",
       "2             MLP  0.696970         0.002369        0.000647\n",
       "3             KNN  0.653680         2.420266        0.000878\n",
       "4             QDA  0.779221         1.620014        0.000993"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "columns = ['Classificador', 'Acurácia', 'Tempo de Treino', 'Tempo de Teste']\n",
    "linha = ['Modelo Shallow', 'Modelo Deep', 'MLP', 'KNN', 'QDA']\n",
    "\n",
    "df_linha = pd.DataFrame(data = linha, columns = ['Classificador'] )\n",
    "df_acu = pd.DataFrame(data = acuracia, columns = ['Acurácia'])\n",
    "df_tempo_de_treino = pd.DataFrame(data = tempo_de_treino, columns = ['Tempo de Treino'])\n",
    "df_tempo_de_teste = pd.DataFrame(data = tempo_de_teste, columns = ['Tempo de Teste'])\n",
    "\n",
    "pd.concat([df_linha, df_acu, df_tempo_de_treino, df_tempo_de_teste], axis = 1, join = 'inner')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "compliant-protection",
   "metadata": {},
   "source": [
    "## *** Comentários sobre o resultado ***"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "corporate-peripheral",
   "metadata": {},
   "source": [
    "Como observado, a maior acurácia foi a do QDA, contudo o melhor modelo foi o MLP, pois sua acurácia chega a quase 70% e o seu tempo de treino foi BEM inferior ao do QDA, o tempo de teste também (apesar de ser bem próximo).É interessante observar que todos os 5 tiveram acurácia próximos de 70%. As diferenças gritantes mesmo foram no tempo de treino. Tempo de treino muito grande é sinônimo de maior custo."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "generous-wages",
   "metadata": {},
   "source": [
    "### *Questão 4 - Considerando o dataset MNIST (importe a partir do keras.datasets), dataset dos dígitos manuscritos utilizado nos exemplos de CNN em aula. Faça o treinamento e avalie as métricas de treino e teste, para os modelos Fully Connected Model e Convolutional Neural Network, vistos em aula e compare os resultados considerando as métricas: Acurácia, Tempo de treino e tempo de teste em relação aos classificadores MLP, KNN e QDA (para estes classificadores, lembre-se de utilizar a vetorização da imagem). Crie uma tabela com as métricas e liste os resultados, comente brevemente os resultados apresentados. Considere a Tabela 1 para listar os resultados.*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "physical-fossil",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from keras.datasets import mnist\n",
    "from keras.utils.np_utils import to_categorical\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Flatten, Activation, Conv2D, MaxPool2D, Dense\n",
    "from tensorflow.keras import backend \n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.optimizers import SGD\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.metrics import accuracy_score\n",
    "import time\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "compound-tracy",
   "metadata": {},
   "outputs": [],
   "source": [
    "(X_train, y_train), (X_test, y_test) = mnist.load_data(path='/tmp/mnist.npz')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "authentic-merchandise",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 28, 28)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "planned-compiler",
   "metadata": {},
   "outputs": [],
   "source": [
    "acuracia = []\n",
    "tempo_de_treino = []\n",
    "tempo_de_teste = []"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "determined-teach",
   "metadata": {},
   "source": [
    "## *** CNN ***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "polished-change",
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_categ = to_categorical(y_train) \n",
    "y_test_categ = to_categorical(y_test)\n",
    "\n",
    "X_train = X_train.reshape(-1, 28, 28, 1)\n",
    "X_test = X_test.reshape(-1, 28, 28, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "least-exception",
   "metadata": {},
   "outputs": [],
   "source": [
    "backend.clear_session()\n",
    "\n",
    "#AM: Feature extraction stages\n",
    "model = Sequential()\n",
    "model.add(Conv2D(64, (3,3), input_shape=(28, 28, 1), padding=('valid')))\n",
    "model.add(MaxPool2D(pool_size=(2,2)))\n",
    "model.add(Activation('relu'))\n",
    "\n",
    "model.add(Flatten())\n",
    "\n",
    "#AM: Fully connected stages (Classification)\n",
    "model.add(Dense(128, activation='relu'))\n",
    "model.add(Dense(10, activation='softmax'))\n",
    "\n",
    "model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "exact-waste",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "329/329 [==============================] - 44s 128ms/step - loss: 18.5273 - accuracy: 0.7987 - val_loss: 0.1471 - val_accuracy: 0.9560\n",
      "Epoch 2/10\n",
      "329/329 [==============================] - 41s 123ms/step - loss: 0.0692 - accuracy: 0.9799 - val_loss: 0.0963 - val_accuracy: 0.9773\n",
      "Epoch 3/10\n",
      "329/329 [==============================] - 41s 125ms/step - loss: 0.0312 - accuracy: 0.9912 - val_loss: 0.0855 - val_accuracy: 0.9791\n",
      "Epoch 4/10\n",
      "329/329 [==============================] - 42s 128ms/step - loss: 0.0138 - accuracy: 0.9960 - val_loss: 0.1088 - val_accuracy: 0.9804\n",
      "Epoch 5/10\n",
      "329/329 [==============================] - 44s 133ms/step - loss: 0.0103 - accuracy: 0.9971 - val_loss: 0.1198 - val_accuracy: 0.9796\n",
      "Epoch 6/10\n",
      "329/329 [==============================] - 44s 133ms/step - loss: 0.0078 - accuracy: 0.9978 - val_loss: 0.1465 - val_accuracy: 0.9817\n",
      "Epoch 7/10\n",
      "329/329 [==============================] - 43s 130ms/step - loss: 0.0056 - accuracy: 0.9983 - val_loss: 0.1875 - val_accuracy: 0.9809\n",
      "Epoch 8/10\n",
      "329/329 [==============================] - 43s 130ms/step - loss: 0.0047 - accuracy: 0.9984 - val_loss: 0.1731 - val_accuracy: 0.9826\n",
      "Epoch 9/10\n",
      "329/329 [==============================] - 45s 135ms/step - loss: 0.0043 - accuracy: 0.9990 - val_loss: 0.1875 - val_accuracy: 0.9824\n",
      "Epoch 10/10\n",
      "329/329 [==============================] - 43s 131ms/step - loss: 0.0025 - accuracy: 0.9994 - val_loss: 0.1852 - val_accuracy: 0.9826\n"
     ]
    }
   ],
   "source": [
    "ti_treino = time.time()\n",
    "a = model.fit(X_train, y_train_categ, batch_size=128, epochs=10, verbose=1, validation_split=0.3)\n",
    "tf_treino = time.time()\n",
    "\n",
    "tempo_de_treino_cnn = tf_treino - ti_treino\n",
    "\n",
    "tempo_de_treino.append(tempo_de_treino_cnn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "reverse-charleston",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "313/313 [==============================] - 4s 12ms/step - loss: 0.1865 - accuracy: 0.9840\n"
     ]
    }
   ],
   "source": [
    "ti_teste = time.time()\n",
    "scores = model.evaluate(X_test, y_test_categ)\n",
    "tf_teste = time.time()\n",
    "\n",
    "tempo_de_teste_cnn = tf_teste - ti_teste\n",
    "\n",
    "tempo_de_teste.append(tempo_de_teste_cnn)\n",
    "acuracia.append(scores[1])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "distinct-visit",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia:  0.984000027179718\n",
      "Tempo de Treino:  429.13174414634705\n",
      "Tempo de Teste:  3.829099655151367\n"
     ]
    }
   ],
   "source": [
    "print('Acurácia: ', scores[1])\n",
    "print('Tempo de Treino: ', tempo_de_treino_cnn)\n",
    "print('Tempo de Teste: ', tempo_de_teste_cnn)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "hawaiian-phone",
   "metadata": {},
   "source": [
    "## *** Fully Connected Model ***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "municipal-defeat",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = X_train.reshape(-1, 28*28)\n",
    "X_test = X_test.reshape(-1, 28*28)\n",
    "\n",
    "X_train = X_train.astype('float32')/255\n",
    "X_test = X_test.astype('float32')/255\n",
    "\n",
    "y_train_categ = to_categorical(y_train) \n",
    "y_test_categ = to_categorical(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "elementary-galaxy",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(512, input_dim=28*28, activation='relu'))\n",
    "model.add(Dense(256, activation='relu'))\n",
    "model.add(Dense(128, activation='relu'))\n",
    "model.add(Dense(32, activation='relu'))\n",
    "model.add(Dense(10, activation='softmax'))\n",
    "\n",
    "model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "multiple-array",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "329/329 [==============================] - 9s 25ms/step - loss: 0.6077 - accuracy: 0.8159 - val_loss: 0.2364 - val_accuracy: 0.9291\n",
      "Epoch 2/10\n",
      "329/329 [==============================] - 8s 24ms/step - loss: 0.1181 - accuracy: 0.9633 - val_loss: 0.1104 - val_accuracy: 0.9678\n",
      "Epoch 3/10\n",
      "329/329 [==============================] - 8s 24ms/step - loss: 0.0683 - accuracy: 0.9792 - val_loss: 0.1270 - val_accuracy: 0.9656\n",
      "Epoch 4/10\n",
      "329/329 [==============================] - 8s 25ms/step - loss: 0.0481 - accuracy: 0.9854 - val_loss: 0.1308 - val_accuracy: 0.9674\n",
      "Epoch 5/10\n",
      "329/329 [==============================] - 8s 26ms/step - loss: 0.0376 - accuracy: 0.9880 - val_loss: 0.1164 - val_accuracy: 0.9719\n",
      "Epoch 6/10\n",
      "329/329 [==============================] - 8s 25ms/step - loss: 0.0273 - accuracy: 0.9919 - val_loss: 0.1230 - val_accuracy: 0.9740\n",
      "Epoch 7/10\n",
      "329/329 [==============================] - 8s 26ms/step - loss: 0.0218 - accuracy: 0.9929 - val_loss: 0.1206 - val_accuracy: 0.9741\n",
      "Epoch 8/10\n",
      "329/329 [==============================] - 9s 26ms/step - loss: 0.0184 - accuracy: 0.9941 - val_loss: 0.1276 - val_accuracy: 0.9754\n",
      "Epoch 9/10\n",
      "329/329 [==============================] - 10s 29ms/step - loss: 0.0157 - accuracy: 0.9957 - val_loss: 0.1339 - val_accuracy: 0.9758\n",
      "Epoch 10/10\n",
      "329/329 [==============================] - 9s 28ms/step - loss: 0.0121 - accuracy: 0.9959 - val_loss: 0.1438 - val_accuracy: 0.9769\n"
     ]
    }
   ],
   "source": [
    "ti_treino = time.time()\n",
    "a = model.fit(X_train, y_train_categ, batch_size=128, epochs=10, verbose=1, validation_split=0.3)\n",
    "tf_treino = time.time()\n",
    "\n",
    "tempo_de_treino_fully = tf_treino - ti_treino\n",
    "tempo_de_treino.append(tempo_de_treino_fully)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "through-fluid",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "313/313 [==============================] - 1s 4ms/step - loss: 0.1209 - accuracy: 0.9799\n"
     ]
    }
   ],
   "source": [
    "ti_teste = time.time()\n",
    "res= model.evaluate(X_test, y_test_categ)\n",
    "tf_teste = time.time()\n",
    "\n",
    "tempo_de_teste_fully = tf_teste - ti_teste\n",
    "\n",
    "tempo_de_teste.append(tempo_de_teste_fully)\n",
    "acuracia.append(res[1])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "banned-unknown",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia:  0.9799000024795532\n",
      "Tempo de Treino:  85.94906854629517\n",
      "Tempo de Teste:  1.4014310836791992\n"
     ]
    }
   ],
   "source": [
    "print('Acurácia: ', res[1])\n",
    "print('Tempo de Treino: ', tempo_de_treino_fully)\n",
    "print('Tempo de Teste: ', tempo_de_teste_fully)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "humanitarian-township",
   "metadata": {},
   "source": [
    "## *** MLP ***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "considerable-textbook",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 1, loss = 1.54722888\n",
      "Iteration 2, loss = 0.69607162\n",
      "Iteration 3, loss = 0.53982427\n",
      "Iteration 4, loss = 0.44773126\n",
      "Iteration 5, loss = 0.38394795\n",
      "Iteration 6, loss = 0.33679740\n",
      "Iteration 7, loss = 0.30145054\n",
      "Iteration 8, loss = 0.27287221\n",
      "Iteration 9, loss = 0.25022272\n",
      "Iteration 10, loss = 0.23078669\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/vivi/.local/lib/python3.8/site-packages/sklearn/neural_network/_multilayer_perceptron.py:614: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (10) reached and the optimization hasn't converged yet.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "mlp = MLPClassifier(solver='adam', learning_rate='adaptive', hidden_layer_sizes = (100,), \n",
    "                        max_iter=10, learning_rate_init=5e-04, tol=1e-4, random_state = 1, verbose = 1)\n",
    "\n",
    "tit_mlp = time.time()\n",
    "mlp.fit(X_train, y_train_categ)\n",
    "tft_mlp = time.time()\n",
    "    \n",
    "ti_mlp = time.time()\n",
    "mlp_predicted = mlp.predict(X_test)\n",
    "tf_mlp = time.time()\n",
    "    \n",
    "    \n",
    "acu_mlp = accuracy_score(y_test_categ, mlp_predicted)\n",
    "acuracia.append(acu_mlp)\n",
    "\n",
    "tempo_de_treino_mlp = tft_mlp - tit_mlp\n",
    "tempo_de_treino.append(tempo_de_treino_mlp)\n",
    "tempo_de_teste_mlp = tf_mlp - ti_mlp\n",
    "tempo_de_teste.append(tempo_de_teste_mlp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "residential-recruitment",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia:  0.9382\n",
      "Tempo de Treino:  24.204227924346924\n",
      "Tempo de Teste:  0.10535883903503418\n"
     ]
    }
   ],
   "source": [
    "print('Acurácia: ', acu_mlp)\n",
    "print('Tempo de Treino: ', tempo_de_treino_mlp)\n",
    "print(\"Tempo de Teste: \", tempo_de_teste_mlp)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "regional-explanation",
   "metadata": {},
   "source": [
    "## *** KNN ***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "ecological-regard",
   "metadata": {},
   "outputs": [],
   "source": [
    "neigh = KNeighborsClassifier(n_neighbors=2)\n",
    "\n",
    "tit_knn = time.time()\n",
    "neigh.fit(X_train, y_train)\n",
    "tft_knn = time.time()\n",
    "\n",
    "ti_knn = time.time()\n",
    "knn_predicted = neigh.predict(X_test)\n",
    "tf_knn = time.time()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "resistant-crowd",
   "metadata": {},
   "outputs": [],
   "source": [
    "acu_knn = accuracy_score(y_test, knn_predicted)\n",
    "acuracia.append(acu_knn)\n",
    "\n",
    "tempo_de_treino_knn = tft_knn - tit_knn\n",
    "tempo_de_treino.append(tempo_de_treino_knn)\n",
    "tempo_de_teste_knn = tf_knn - ti_knn\n",
    "tempo_de_teste.append(tempo_de_teste_knn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "improving-nowhere",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia:  0.9627\n",
      "Tempo de Treino:  0.11896038055419922\n",
      "Tempo de Teste:  125.49846482276917\n"
     ]
    }
   ],
   "source": [
    "print('Acurácia: ', acu_knn)\n",
    "print('Tempo de Treino: ', tempo_de_treino_knn)\n",
    "print(\"Tempo de Teste: \", tempo_de_teste_knn)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "medieval-scanning",
   "metadata": {},
   "source": [
    "## *** QDA ***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "checked-underwear",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/vivi/.local/lib/python3.8/site-packages/sklearn/discriminant_analysis.py:808: UserWarning: Variables are collinear\n",
      "  warnings.warn(\"Variables are collinear\")\n"
     ]
    }
   ],
   "source": [
    "qda = QuadraticDiscriminantAnalysis()\n",
    "tit_qda = time.time()\n",
    "qda.fit(X_train, y_train)\n",
    "tft_qda = time.time()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "protecting-office",
   "metadata": {},
   "outputs": [],
   "source": [
    "ti_qda = time.time()\n",
    "qda_predicted = qda.predict(X_test) \n",
    "tf_qda = time.time()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "native-directory",
   "metadata": {},
   "outputs": [],
   "source": [
    "acu_qda = accuracy_score(y_test, qda_predicted)\n",
    "acuracia.append(acu_qda)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "accomplished-batch",
   "metadata": {},
   "outputs": [],
   "source": [
    "tempo_de_treino_qda = tft_qda - tit_qda\n",
    "tempo_de_treino.append(tempo_de_treino_qda)\n",
    "tempo_de_teste_qda = tf_qda - ti_qda\n",
    "tempo_de_teste.append(tempo_de_teste_qda)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "finnish-voltage",
   "metadata": {},
   "source": [
    "# *Tabela com as Métricas (listar os resultados)*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "julian-conflict",
   "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>Classificador</th>\n",
       "      <th>Acurácia</th>\n",
       "      <th>Tempo de Treino</th>\n",
       "      <th>Tempo de Teste</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>CNN</td>\n",
       "      <td>0.9627</td>\n",
       "      <td>0.118960</td>\n",
       "      <td>125.498465</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Fully Connected Model</td>\n",
       "      <td>0.1403</td>\n",
       "      <td>33.634007</td>\n",
       "      <td>3.836388</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           Classificador  Acurácia  Tempo de Treino  Tempo de Teste\n",
       "0                    CNN    0.9627         0.118960      125.498465\n",
       "1  Fully Connected Model    0.1403        33.634007        3.836388"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "columns = ['Classificador', 'Acurácia', 'Tempo de Treino', 'Tempo de Teste']\n",
    "linha = ['CNN', 'Fully Connected Model', 'MLP', 'KNN', 'QDA']\n",
    "\n",
    "df_linha = pd.DataFrame(data = linha, columns = ['Classificador'] )\n",
    "df_acu = pd.DataFrame(data = acuracia, columns = ['Acurácia'])\n",
    "df_tempo_de_treino = pd.DataFrame(data = tempo_de_treino, columns = ['Tempo de Treino'])\n",
    "df_tempo_de_teste = pd.DataFrame(data = tempo_de_teste, columns = ['Tempo de Teste'])\n",
    "\n",
    "pd.concat([df_linha, df_acu, df_tempo_de_treino, df_tempo_de_teste], axis = 1, join = 'inner')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "resistant-andorra",
   "metadata": {},
   "source": [
    "## *** Comentários sobre o resultado ***"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "perceived-dallas",
   "metadata": {},
   "source": [
    "Obs.: Aqui eu não sei o porquê, mas só visualizo o CNN e o fully, o restante não está aparecendo na tabela (eu revisei, mas não enxerguei erro e como já estava próximo a entrega, eu não tive como pensar mais no que houve)\n",
    "\n",
    "COm o que eu pude ver aqui, o que melhor se saiu foi o CNN. Existe uma diferença absurda entre ele e o FUlly. A Acurácia do CNN é altíssima, enquanto a do fully é baixíssima. O tempo de treino do fully é muito longo, enquanto o do CNN é totalmente ao contrário. A única característica que foi do contra nos dois foi o tempo de teste, que a situação se inverteu. No mais, o melhor que se saiu foi o CNN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "durable-precipitation",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}