Comparison of interval lengths for the intuitionistic fuzzy time series forecasting model / Nik Muhammad Farhan Hakim Nik Badrul Alam … [et al.]

A fuzzy time series forecasting model can cater for the time series data described by linguistic terms. The use of fuzzy sets in forecasting time series data is evidently better at predicting data compared to the classical time series model. The fuzzy set concept was extended to the intuitionistic fuzzy set, in which its performance in forecasting time series data is extensively better than the classical fuzzy set. In the intuitionistic fuzzy time series forecasting model, the universe of discourse is defined and divided into several intervals before the data are fuzzified. The objective of this study is to compare the forecasting performance using different interval lengths. The historical data of student enrollments at the University of Alabama were adopted, in which 7, 14, and 21 intervals were used to perform the forecasting process. The results have shown that the model with 21 sub-intervals outperformed the other models. In the future, it is recommended that researchers determine an effective interval length at the early stage of forecasting to obtain the best performance result for time series forecasting.