Comparison of intuitionistic fuzzy time series forecasting models using different interval partitioning methods in predicting Malaysian crude palm oil prices

• Published in: AIP Conference Proceedings
• Main Author: Alam N.M.F.H.N.B.; Ramli N.; Nassir A.A.; Mohd A.H.; Mohammed N.
• Format: Conference paper
• Language: English
• Published: American Institute of Physics 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85188460861&doi=10.1063%2f5.0192601&partnerID=40&md5=e06a3d6aaae952345513e4db67392430

An intuitionistic fuzzy time series forecasting (IFTSF) model is capable of handling the non-determinism in time series data. Partitioning the universe of discourse into several intervals is one of the preliminary steps in conducting the IFTSF. The effective interval length for the IFTSF needs to be determined to decrease the computing complexity and speed up the forecasting procedure. This research develops the IFTSF model using three different interval partitioning methods, namely the average-based (L1), the frequency-density-based (L2) and the redivide-randomly chosen length (L3). The data are fuzzified using triangular fuzzy numbers, and the intuitionistic fuzzy sets (IFS) are then built. The IFS is then defuzzified using the crispification formula, which preserves the nature of the IFS. The Malaysian crude palm oil prices data are used to illustrate the proposed model. The forecasting accuracy of each model is measured, and the results show that the redivide-randomly chosen length outperforms the other interval partitioning methods. © 2024 Author(s).