wine_mlp_deep_learning.py

# -*- coding: utf-8 -*-
"""Wine-MLP - Deep Learning - Atividade Aba4 .ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1Q8AmXCH7SioOXEF0LxZcYggyUSyPaka_
"""

import pandas as pd
import tensorflow
from tensorflow import keras
from keras.models import model_from_json
import matplotlib.pyplot as plt
import numpy as np
from sklearn.model_selection import train_test_split

base = pd.read_csv('wine.csv')

entradas = base[base.columns[1:]].to_numpy()
saidas = base[base.columns[0]]

valoresSaida = np.empty((177, 1), dtype=int)

for i in range(177):
   valoresSaida[i]= saidas[i]-1

entradasTrain, entradasTest, valoresSaidaTrain, valoresSaidaTest = train_test_split(entradas, valoresSaida, test_size=0.3)

print("Quantidade de Dados de Treino:" ,len(entradasTrain))
print("Quantidade de Dados de Teste:" ,len(entradasTest))

print("Quantidade de Dados de Treino e dos atributos:" ,entradasTrain.shape)
print("Quantidade de Dados de Teste e dos atributos" ,entradasTest.shape)

print("Quantidade de Saidas de Treino e coluna:" ,valoresSaidaTrain.shape)
print("Quantidade de Saidas de Teste e coluna:" ,valoresSaidaTest.shape)

print("min: ", valoresSaidaTrain.min())
print("max: ", valoresSaidaTrain.max())

modelo = keras.Sequential([keras.layers.Dropout(0.2),
                           keras.layers.Dense(130, activation=tensorflow.nn.relu),
                           keras.layers.Dense(70, activation=tensorflow.nn.relu),
                           keras.layers.Dense(40, activation=tensorflow.nn.relu),
                           keras.layers.Dense(13, activation=tensorflow.nn.relu),
                           keras.layers.Dense(3, activation=tensorflow.nn.softmax)
                           ])

modelo.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics='accuracy')

hist=modelo.fit(entradasTrain, valoresSaidaTrain, epochs=70, validation_split=0.3)

plt.plot(hist.history['accuracy'])
plt.plot(hist.history['val_accuracy'])
plt.title('Acurácia por épocas')
plt.xlabel('Épocas')
plt.ylabel('Acurácia')
plt.legend(['Treino', 'Validação'])

plt.plot(hist.history['loss'])
plt.plot(hist.history['val_loss'])
plt.title('Loss por épocas')
plt.xlabel('Épocas')
plt.ylabel('Loss')
plt.legend(['Treino', 'Validação'])

model_json = modelo.to_json()
with open("model.json", "w") as json_file:
   json_file.write(model_json)

modelo.save_weights("model.h5")
print("Modelo Salvo")

json_file = open("model.json", "r")
loaded_model_json = json_file.read()
json_file.close()

loaded_model = model_from_json(loaded_model_json)
loaded_model.load_weights("model.h5")
print("Modelo Carregado")

loaded_model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics='accuracy')

perda_teste, acuracia_teste = loaded_model.evaluate(entradasTest, valoresSaidaTest)
print("Perda do teste: ", perda_teste)
print("Acurácia do teste: ", acuracia_teste)
testes=loaded_model.predict(entradasTest)

x=0
for i in testes:
   print(np.argmax(i), valoresSaidaTest[x])
   x=x+1