SET Index Forecast Using Bayesian Belief Networks

摘要

Statistical time-series forecasting faces the problem of accuracy if the data deviate from a normal distribution. In literature, stock indexes are not of normal distribution. This paper presents the Bayesian Belief Network (BBN) in forecasting the Stock Exchange of Thailand (SET Index) in comparison with statistical forecasting techniques. To model BBN, SET index distribution is discretized using a number of clustering techniques for comparison. Then, BBN is constructed using the transforming data in a P/E ratio via the K2 algorithm based on the training dataset gathered from January 2013 through July 2019. For performance evaluation, the proposed model was compared with the statistical forecasting algorithms using RMSE and the correlation coefficient (CC). The results show that the proposed BBN with a particular clustering algorithm provided better results than the statistical forecasting techniques.