Compound Interest Long Short-Term Memory for Stock Return Forecast

Great effort has been made to predict the price, return or trend of stock, but it is still hard to forecast them due to the unpredictable factors in stock market. With the rapid development of deep learning, experts begin to apply deep learning to financial problems. In this paper, we propose Compound Interest Long Short-Term Memory (CILSTM) for stock return forecast. It is of great importance to make a profit from compound interest in stock market. Therefore, we introduce compound interest to Long Short-Term Memory (LSTM) to capture the features in investing stock. Moreover, ci(n) function is designed to the cell of CILSTM to describe compound interest quantitatively. 50 stocks in Chinese stock market were selected as the dataset to evaluate the models. The experiment demonstrates more competitive performance by the CILSTM as compared with the state-of-the-art methods.