Stock index forecasting: A new fuzzy time series forecasting method

Abstract

Abstract This paper presents a new fuzzy time series forecasting model based on technical analysis, affinity propagation (AP) clustering, and a support vector regression (SVR) model. Technical analysis indicators are divided into three categories to construct multivariate fuzzy logical relationships. AP clustering without specifying the number of clusters is used to obtain a suitable partition for the universe of discourse, and the representative exemplars are generated as defuzzied values. The SVR model is employed to explore the unrecognized relationships and modify the forecasts. In addition, the error‐based evaluation criteria are applied to evaluate the methods. The performance of the method is evaluated using the Taiwan Capitalization Weighted Stock Index (TAIEX), Standard & Poor's 500 Index (S&P500), and Dow Jones Industrial Average (DJIA) dataset, and the experimental results demonstrate that the proposed method outperforms some classic models.