x_new_batch = scaled_data[-back_window:, :-1]
x_new_batch = x_new_batch.reshape(1, x_new_batch.shape[0],
x_new_batch.shape[1])
pred = model.predict(x_new_batch)
temp = [0] * new_batch.shape[1]
temp[-1] = pred
real_value_pred = batch_scale.inverse_transform([temp])[0][-1]
return real_value_pred
if __name__ == '__main__':
lstm = LSTM()
df = pd.read_csv('000002-from-1995-01-01.csv')
window = 20
X_train, y_train, X_test, y_test = lstm.preprocess_data(df[::-1],
window,
predict_length=1,
split_percent=0.85)
model = lstm.build_model([X_train.shape[2], window, 100, 1],
dropout=0.3,
problem_class='classification')
encoder = LabelEncoder()
encoded_Y = encoder.fit_transform(y_train)
dummy_y = np_utils.to_categorical(encoded_Y)
model.fit(X_train,
dummy_y,
batch_size=768,
nb_epoch=10,
validation_split=0.1,
verbose=1)
