2019/01/20 by Kei Nakagawa, Nakagawa, Kei, Tomoki Ito +5
Computer Science · Decision Sciences · Economics, Econometrics and Finance · Engineering · #Energy Load and Power Forecasting #FOS: Computer and information sciences #FOS: Economics and business #Financial Markets and Investment Strategies #Machine Learning (cs.LG) #Neural Networks and Applications #Statistical Finance (q-fin.ST) #Stock Market Forecasting Methods
paper · pdf · doi:10.48550/arxiv.1901.11493
openalex publication_date 2019/01/20 · openalex created_date 2021/02/01 · openalex updated_date 2026/07/28
A linear multi-factor model is one of the most important tools in equity\nportfolio management. The linear multi-factor models are widely used because\nthey can be easily interpreted. However, financial markets are not linear and\ntheir accuracy is limited. Recently, deep learning methods were proposed to\npredict stock return in terms of the multi-factor model. Although these methods\nperform quite well, they have significant disadvantages such as a lack of\ntransparency and limitations in the interpretability of the prediction. It is\nthus difficult for institutional investors to use black-box-type machine\nlearning techniques in actual investment practice because they should show\naccountability to their customers. Consequently, the solution we propose is\nbased on LSTM with LRP. Specifically, we extend the linear multi-factor model\nto be non-linear and time-varying with LSTM. Then, we approximate and linearize\nthe learned LSTM models by LRP. We call this LSTM+LRP model a deep recurrent\nfactor model. Finally, we perform an empirical analysis of the Japanese stock\nmarket and show that our recurrent model has better predictive capability than\nthe traditional linear model and fully-connected deep learning methods.\n