Bidirectional(GRU(units=50, return_sequences=True)),
tfa.layers.GroupNormalization(),
Dropout(0.2),
Bidirectional(GRU(units=50, return_sequences=True)),
tfa.layers.GroupNormalization(),
Dropout(0.2),
Bidirectional(LSTM(units=50)),
tfa.layers.GroupNormalization(),
Dropout(0.2),
Dense(units=1, activation='sigmoid')
]
name = 'bgru_lstm'
epochs = 100
batch_size = 32
# Training the model.
model = rnn_model.build_model(input_shape, layers)
model, history = rnn_model.train(model,
name,
x_train,
y_train,
epochs=epochs,
batch_size=batch_size)
# Making predictions.
y_predict = model.predict(x_test)
# Visualising the prediction.
stocks.plot_prediction(y_test, y_predict, 'Real Tesla Stock Price',
'Predicted Tesla Stock Price')
_, test_data = stocks.read_data(preprocessing=True, scaling_method='MinMax')
x_test, y_test = stocks.construct_time_frames(test_data, frame_size=16)
# Defining layers of LSTM.
input_shape=x_test.shape[1:]
layers = [
Bidirectional(GRU(units=50, return_sequences=True)),
tfa.layers.GroupNormalization(),
Dropout(0.2),
Bidirectional(GRU(units=50, return_sequences=True)),
tfa.layers.GroupNormalization(),
Dropout(0.2),
Bidirectional(LSTM(units=50)),
tfa.layers.GroupNormalization(),
Dropout(0.2),
Dense(units=1, activation='sigmoid')
]
# Making predictions.
model = rnn_model.build_model(input_shape, layers)
model.load_weights('weights/bgru_lstm_callback.h5')
y_predict = model.predict(x_test)
# Visualising the prediction.
stocks.plot_prediction(y_test, y_predict, 'Real ALPHA BANK Stock Price', 'Predicted ALPHA BANK Stock Price')
for i, y_pred in enumerate(y_predict):
print('Day', i, ':', y_pred[0])