“Darling, time horizon matters”: applying the CNN-LSTM method for predicting US equity ETFs

The paper uses a hybrid model of convolutional neural network and long short-term memory (CNN-LSTM) to examine the impact of the prediction (or input) window length on the prediction accuracy of trading decisions for US equity ETFs. Results demonstrate that the prediction window length plays a vital role in the prediction accuracy of the trading decision. A hump-shaped relationship is observed between prediction accuracy and prediction window length over monthly trading days.