Using Long Short-Term Memory to Predict Cash Dividend

Abstract

Investing in stocks has been very popular in recent years. Investors hope to make a profit by investing in stocks. However, stock prices are highly volatile. Many investors judge whether to invest in stocks based on historical stock prices, technical indicators and market conditions. There are many researches about stock price forecasting in the research field but seldom for cash dividend. The profitable methods of investing in stocks include earning the stock spread and the cash dividends that the company distributes to shareholders. However, cash dividends are major profit for the long-term investor. In this research, we proposed a Stacked Long Short-Term Memory model to predict the value of the company's annual distribution of cash dividends that provided a basis for managers and investors to make investment decisions. The experiments showed the accuracy of prediction for Cash Dividend of the Formosa Plastics Corporation is excellent that the MAPE is only 2.95%.