A computational method based on interval length for fuzzy time series forecasting

Abstract

In the literature, there have been a good many different forecasting methods related to forecasting problems of fuzzy time series. The main issue of fuzzy time series forecasting is the accuracy of the forecasted values. The forecasting accuracy rate is affected by the length of each interval in the universe of discourse. Thus, it is substantial to determine the length of each interval. In this study, a new computational method based on class width to determine interval length is proposed and also used the coefficient of variation for time series forecasting. After the intervals are formed, the historical time series data set is fuzzified according to fuzzy time series theory. The proposed model has been tested on the student enrollments, University of Alabama, and a real-life problem of rice production for containing higher uncertainty. This method was compared with existent methods to determine the effectiveness in terms of the mean square error (MSE) and the average forecasting (AFE). The results are shown that the proposed model can achieve a higher forecasting accuracy rate than the existing models.