Computational-based partitioning and Strong [formula omitted]-cut based novel method for intuitionistic fuzzy time series forecasting

Abstract

In this article, we propose a computational-based partitioning (CBP) and Strong α,β-cut based novel intuitionistic fuzzy time series (IFTS) forecasting method. Construction of intervals, intuitionistic fuzzification of time series data, appropriate intuitionistic fuzzy logical relationships (IFLRs), and procedure of defuzzification are critical issues that affect the forecasting accuracy of any IFTS method. A Computational-based partitioning approach uses basic statistical parameters to determine the number of intervals and constructing of intervals without specialized knowledge of the domain. For intuitionistic fuzzification of time series data, all those intuitionistic fuzzy sets are taken having both non-zero membership and non-membership grade. A Strong α,β-cut are used to choose apposite IFLRs that deliver importance in investigating the tendency of time series data. We also propose a defuzzification approach to get the numerical values. In this article, two popular historical time series datasets are used to demonstrate the supremacy of the proposed forecasting method. Root mean square error (RMSE) and mean absolute percentage error (MAPE) are used to verify the performance of the proposed method. Verification of the validity of the proposed forecasting method is authenticated by using the values of the Theil inequality coefficient (U) and tracking signal (TS).