n=nclasses,
normalized=normalize_class)
print('-' * 20)
print('\nMATTHEWS CORRELATION')
print(matthews_corrcoef(test_y, test_yy))
CM = confusion_matrix(test_y, test_yy, labels=[1, 0, -1])
print('\nCONFUSION MATRIX')
print(CM / CM.astype(np.float).sum(axis=1))
print('\nCLASSIFICATION REPORT')
print(classification_report(test_y, test_yy, labels=[1, 0, -1], target_names=['buy', 'hold', 'sell']))
print('-' * 20)
plt.plot(predicted)
plt.title('Predict')
plt.ylabel('class')
plt.xlabel('bar')
plt.legend(['buy', 'hold', 'sell'])
plt.show()
plt.plot(data_yy)
plt.title('Saved predict')
plt.ylabel('class')
plt.xlabel('bar')
plt.legend(['prediction'])
plt.show()
if run_type == 0:
plot_history(history)
# Train
history = autoencoder.fit(
x_data,
x_data,
epochs=epochs,
batch_size=batch,
validation_split=0.1,
shuffle=True,
# callbacks=[tb]
)
# Save weights
autoencoder.save_weights(wgt_path + model_name + '.hdf5')
encoder.save_weights(wgt_path + model_name + '_enc.hdf5')
plot_history(history, acc='acc')
## Visual evaluation
# Number of examples = n^2
n = 10
test_z_mean, test_z_log_var, test_z = encoder.predict(x_data[-100:],
batch_size=batch)
# Save images of origin and decoded data
if action in ['train1', 'train2', 'predict']:
def plot_data(fname, plot_data):
figure = np.zeros((img_width * n, img_width * n, 3))
plot_data = plot_data[-(n * n):]
for idx in range(n):
for jdx in range(n):
np.savetxt(file_yy, data_yy, fmt='%.2f', delimiter=',')
print("Predict saved:\n", file_yy)
#=============================================================================#
# P L O T #
#=============================================================================#
# test_y_ = [np.argmax(x) for x in test_y]
# test_yy = [np.argmax(x) for x in model.predict(test_x).reshape(test_x.shape[0], nclasses)]
# true, test
print(classification_scores(data_yyy, data_yy, nclasses))
# plt.plot(predicted)
# plt.title('Predict')
# plt.ylabel('class')
# plt.xlabel('bar')
# plt.legend(['buy', 'hold', 'sell'])
# plt.show()
plt.plot(data_yy)
plt.title('Saved predict')
plt.ylabel('class')
plt.xlabel('bar')
plt.legend(['prediction'])
plt.show()
plot_history(history, acc='categorical_accuracy')
