Forecasting regime switches to assist decision making

Purpose: This paper aims to propose a novel model to forecast regime switches in a time series to assist decision making. Design/methodology/approach: The authors apply the clustering technique to group the data into five states. Then, a model is proposed to formulate the relationships from in‐sample observations, including regime switch relationships. Afterwards, the model uses the relationships to forecast the regime switches in out‐sample observations. Findings: The study uses daily Taiwan Stock Exchange Capitalization Weighted Stock Index as the forecasting target. Regime switches in in‐sample observations are identified. And a regime switch is successfully forecasted by the proposed model. Research limitations/implications: The proposed model identifies a regime switch which matches the real event. It implies that the proposed model can be applied to other time series, such as Dow Jones or NASDAQ. Originality/value: Previous studies contribute to the forecasting of regime switches. The forecasting results are validated with the real event. One of the forecasted regime switches matches the event of Lehman Brothers' declaring of bankruptcy.