Two-factor high-order fuzzy-trend FTS model based on BSO-FCM and improved KA for TAIEX stock forecasting

Fuzzy time series has been an effective and attractive forecasting model for solving the problem of stock index forecasting. In particular, fuzzy-trend fuzzy time series models have been proposed recently to address complex cases and perform well in terms of forecasting accuracy. Nonetheless, they have just explored the two-factor second-order forecasting but cannot satisfy the complex stock system. In this study, we proposed a new two-factor high-order fuzzy-trend fuzzy time series model to explore the more complex situation on the TAIEX stock index forecasting. We presented the backtracking search optimization-fuzzy c-means method to obtain the optimal intervals of the data sets. In addition, an improved kidney-inspired algorithm is employed to integrate the high-order forecasting values. The proposed model shows outstanding forecasting accuracy than the benchmark methods on the TAIEX. Besides, we combined two other stock indexes (NASDAQ and the Dow Jones) as the secondary factors, respectively. It provides a useful method for two-factor high-order fuzzy-trend fuzzy time series stock index forecasting.