Combining machine learning algorithm with arima for stock market forecasting: The case of set100 index

Abstract

At present, the number of investors in the Stock Exchange of Thailand has continuously increased while the loss of investors also increased due to lack of experience, and they are unable to predict the stock price accurately. This paper proposes a two-stage forecasting model that incorporates a machine learning algorithm such as a decision tree model and parametric techniques such as autoregressive integrated moving average (ARIMA) and aims to improve stock price forecasting. In this case, the decision tree model determines the investment attractiveness of the SET100 Index listed in the Stock Exchange of Thailand, and the group of stocks with high investment potential is identified with 90.48 percent accuracy. According to the decision tree model, the BTS Group Holdings Public Company Limited was chosen from the high investment potential group to predict the short-term closing price trend with the ARIMA model. The ARIMA model can predict precisely with a slight error (p-value < 0.01). Therefore, it can be concluded that the ensemble machine learning methods together with ARIMA can be used as a hybrid method to increase prediction capability for supporting investment decisions.