This paper proposes a novel fuzzy forecasting method for forecasting the TAIEX based on optimal intervals and a similarity measure between the subscripts of fuzzy sets, where two threshold values α and β and a weighting constant γ are used, α ∊ [0, 1], β ∊ [0, 1] and γ ∊ [0, 1]. First, we use particle swarm optimization (PSO) techniques to obtain optimal intervals in the universe of discourses (UODs) of the main factor (MF) TAIEX and the secondary factor (SF), respectively, where SF ∊ {Dow Jones, NASDAQ, M1B}. Based on optimal intervals, the constructed two-factors second-order fuzzy-trend logical relationship groups (TSFTLRGs), and a similarity measure between the subscripts of fuzzy sets, we propose a novel method to deal with the forecasting of the TAIEX. The fuzzy forecasting method presented in this paper gets better performance than the existing methods for dealing with the forecasting of the TAIEX.