A two-factor linguistic time series model with time-dependent semantic relationship groups

Abstract

Fuzzy time series-based forecasting models have been extensively examined because of their flexibility and ability to handle uncertainty. In those forecasting models, the universe of discourse is apportioned into subintervals and human experts intuitively assign fuzzy sets to them. Because fuzzy sets are always associated with linguistic words, the concept of linguistic time series (LTS) is introduced, which a numeric time series is transformed into a linguistic one by a mathematical formalism. Then, the obtained LTS specifies semantic logical relationships that are utilized to generate semantic logical relationship groups (SLRG) for establishing forecasting models. In this paper, the concept of time-dependent multi-factor SLRG is proposed and it is used to establish time-dependent multi-factor first-order and high-order LTS models. The experimental studies executed on the time series datasets of the TAIFEX index of Taiwan and daily average temperature in 1996 in Taipei, Taiwan, show that our proposed forecasting model outperforms the benchmarked models.