Fuzzy time series forecasting based on overlapping partitions
Abhijit Gogoi, Bhogeswar Borah · 2024
This study introduces a novel approach to Fuzzy Time Series (FTS) forecasting which utilizes overlapping partitions of the Universe of Discourse (UOD). The methodology involves creating Fuzzy Logical Relations (FLRs) derived from the overlapping partitions, both at first and high-orders. Real-world datasets including enrollment data from the University of Alabama, market indices, car sales figures, gold prices, lynx and sunspot counts are used for assessment. The performance of the proposed FTS forecasting system is evaluated using Root Mean Square Error (RMSE), consistently demonstrating superior results compared to other approaches, indicating its potential as an effective forecasting method across various domains.