A Proposal of a Method to Determine the Appropriate Learning Period in Stock Price Prediction Using Machine Learning

In this study, we propose a method to determine the appropriate learning period for each stock and the prediction period by considering stock price fluctuations for stock price prediction using machine learning. Our proposed method uses historical volatility as an indicator of the turning point to determine the learning period based on the policy that the fluctuations in the period after the major turning point of stock price fluctuations and the fluctuations in the prediction period are likely to be similar and that the prediction accuracy can be improved by eliminating the period before the turning point from the learning period. We used Long Short-Term Memory (LSTM), which has been used in many related studies on stock price prediction, as the machine learning model. Experiments showed that the accuracy of predictions by neural networks trained with the learning period determined by the proposed method was better than that of predictions by neural networks trained with the same learning period for all stocks. (c) 2024 Institute of Electrical Engineer of Japan and Wiley Periodicals LLC.