Article Published In Vol.9 (March-April 2021)

An Interval Partitioning Approach Using Graph–Based Clustering in Fuzzy Time Series Forecasting Model

Author : Nghiem Van Tinh, Tran Thi Thanh, Bui Thi Thi

In the implementations of fuzzy time series (FTS) forecasting model, the determination of interval lengths has an important impact on the performance of the forecasting model. However, Equal length intervals used in most existing literatures are convenient but have been chosen arbitrarily. Huarng developed a new approach which is called distribution- and average-based length in order to determine effective length of partitioned intervals. In this study, a new FTS forecasting model which uses a graph- based clustering technique to determine different length of intervals is proposed. The proposed forecasting model has been applied to the two time series data, which are well-known enrolments data at The University of Alabama and numerical data sets of Gasonline prices in VietNam. The computational results show that the proposed model gets a higher forecasting accuracy than the existing models when is applied to enrolments data at The University of Alabama.

Keyword.  Enrolments, forecasting, FTS, fuzzy relationships, clustering technique