Natural partitioning-based forecasting model for fuzzy time-series

Since the forecasting framework of fuzzy time-series introduced, there have been a variety of models developed to improve forecasting accuracy or reduce computation overhead. However, the issue of partitioning intervals has rarely been investigated. This work presents a novel approach to handling the issue by applying the natural partitioning technique, which can recursively partition the universe of discourse level by level in a natural way. Experimental results on the enrollment data of the University of Alabama demonstrate that the resulting forecasting model can forecast the data effectively and efficiently and outperforms the existing models. Furthermore, the proposed model can be extended to handle high-order fuzzy time series.