def create_model(x_train_ts, y_train_ts, x_test_ts, y_test_ts):
    x_train_ts, y_train_ts, x_test_ts, y_test_ts = build_data()
    lstm_model = Sequential()
    # (batch_size, timesteps, data_dim)
    lstm_model.add(LSTM({{choice([50, 100, 150])}}, batch_input_shape=(BATCH_SIZE, TIME_STEPS, x_train_ts.shape[2]), dropout=0.2,
                        recurrent_dropout=0.2, stateful=True, return_sequences=True,
                        kernel_initializer='random_uniform'))
    if {{choice(['one_lstm','two_lstm'])}} == 'two_lstm':
        lstm_model.add(LSTM({{choice([30, 60, 80])}}, dropout={{choice([0.1,0.2,0.3])}}))
    if {{choice(['one_dense','two_dense'])}} == 'two_dense':
        lstm_model.add(Dense({{choice([10, 20])}}, activation='relu'))
    lstm_model.add(Dense(1, activation='sigmoid'))
    if {{choice(['sgd','rms'])}} == 'rms':
        optimizer = optimizers.RMSprop(lr={{uniform(000.1, 0.1)}})
    else:
        optimizer = optimizers.SGD(lr={{uniform(000.1, 0.1)}}, decay=1e-6, momentum=0.9, nesterov=True)
    lstm_model.compile(loss='mean_squared_error', optimizer=optimizer)  # binary_crossentropy

    history = lstm_model.fit(x_train_ts, y_train_ts, epochs={{choice([20, 40, 60, 70])}}, verbose=2, batch_size=BATCH_SIZE,
                             validation_data=[x_test_ts, y_test_ts],
                             callbacks=[LogMetrics(search_params, params, comb_no), csv_logger])

    val_error = np.amax(history.history['val_acc'])
    print('Best validation acc of epoch:', val_error)
    return {'loss': val_error, 'status': STATUS_OK, 'model': lstm_model} # if accuracy use '-' sign


best_run, best_model = optim.minimize(model=create_model,
                                      data=data_dummy,
                                      algo=tpe.suggest,
                                      max_evals=2000,
                                      trials=Trials())