EEG-MLP.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Mon Oct 21 00:16:21 2019

@author: jeremiah
"""

#%% importings
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

from keras.models import Sequential
from keras.layers import Dense, Dropout
#from keras.wrappers.scikit_learn import KerasClassifier
from keras.optimizers import SGD
#from keras.constraints import maxnorm
from keras.callbacks import ModelCheckpoint

from scipy.io import arff

#from sklearn.model_selection import cross_val_score, StratifiedKFold
from sklearn.preprocessing import LabelEncoder, StandardScaler
#from sklearn.pipeline import Pipeline
#%%
seed = 7
np.random.seed(7)
#%% data preparation
path = '/home/jeremiah/data_analysis/datasets/EEG-Eye-State.arff'

data = arff.loadarff(path)
df = pd.DataFrame(data[0])
dataset = df.to_numpy().astype(float)

np.random.shuffle(dataset)

test_per = 0.3
train_size = int(len(dataset)*(1-test_per))
X = dataset[:train_size,:-1]
Y = dataset[:train_size, -1]

X_test = dataset[train_size:,:-1]
Y_test = dataset[train_size:,-1]
#%% defining model
def create_model():
    #dp_rate = 0.1


    model = Sequential()
    model.add(Dense(20, input_dim=14, kernel_initializer='normal', activation='relu'))
    #model.add(Dropout(rate=dp_rate))

    #model.add(Dense(32, kernel_initializer='normal', activation='relu',
              #kernel_constraint=maxnorm(3)))
    #model.add(Dropout(rate=dp_rate))

    model.add(Dense(10, kernel_initializer='normal', activation='relu'))
    #model.add(Dropout(rate=dp_rate))

    model.add(Dense(1, kernel_initializer='normal', activation='sigmoid'))

    sgd = SGD(lr=0.065, decay=0.002)

    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
    return model

#%% training model
batch_size = 20
epochs = 300
model = create_model()

scaler = StandardScaler()
scaler.fit(X)
std_x = scaler.transform(X)

checkpoint = ModelCheckpoint('/home/jeremiah/trained_models/best.model.h5',
                             monitor='val_acc', verbose=1, save_best_only=True, mode='max')

callback_list = [checkpoint]

history = model.fit(std_x, Y, validation_split=0.33, batch_size=batch_size, epochs=epochs, verbose=1,
                    callbacks=callback_list)


#%% viewing results
print(history.history.keys())
print('acc: %.2f%%\tval_acc: %.2f%%' %(history.history['acc'][-1]*100, history.history['val_acc'][-1]*100))



plt.figure(figsize=(14, 5))

plt.subplot(1, 2, 1)
plt.plot(history.history['acc'], 'b-', history.history['val_acc'], 'r-')
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='best')

plt.subplot(1, 2, 2)
plt.plot(history.history['loss'], 'b-', history.history['val_loss'], 'r-')
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='best')
plt.savefig('/home/jeremiah/trained_models/EEG-MLP-02.png')

#%% loading best model weights
model.load_weights('/home/jeremiah/trained_models/best.model.h5')
print('model loaded!')

#%% saving model
model_json = model.to_json()
with open('/home/jeremiah/trained_models/EEG-MLP-02.json', 'w') as file:
    file.write(model_json)
model.save_weights('/home/jeremiah/trained_models/EEG-MLP-02.h5')
print('model saved successfully!')

#%%
scaler.fit(X_test)
std_x_test = scaler.transform(X_test)
scores = model.evaluate(std_x_test, Y_test, batch_size=16)
print('Results:\n%s: %.2f%%' %(model.metrics_names[1], scores[1]*100))