A 2-Tuple Fuzzy Time Series and Fuzzy Modeling Inference Method

Abstract

This study proposes a 2-tuple fuzzy time series forecasting method. The 2-tuple linguistic representation was first proposed by Herrera and Martínez [14], and was mainly applied for computing with words, decision making and decision analysis problems. This study applies the 2-tuple concept to the fuzzy time series, and the historical data is fuzzified to 2-tuple linguistic representation. The advantage of 2-tuple linguistic representation is that the data can be fuzzified more accurately and the loss of information is reduced. In addition, the 2-tuple fuzzy logical relationship can be established. A new fuzzy inference process based on proportional shift is also proposed. Finally, this study applies the 2-tuple fuzzy time series method to forecast the enrollments at the University of Alabama. Compared to several other methods, the forecasting results of the present method show better accuracy. For verification, another dataset is also used. This study applies the present method to forecast Taiwan tourists to Europe. Compared to Chen's method[4], the forecasting results of the present method show better accuracy.