Related Theory’s Research Of Fuzzy Time Series Models

The study of fuzzy time series has increasingly attracted much attention due to its salient capabilities of tackling uncertainty and vagueness inherent in data collected. Studies find universe of discourse partition and data fuzzification are two particularly important parties on the fuzzy time series models. At present, the existing fuzzy time series forecasting models still exist the shortcomings in the universe of discourse partition and data fuzzification, the, above two problem will be studied in this paper.In the universe of discourse partition, present studies show that it is very important to partition universe of discourse reasonably according to the sample data distribution characteristics in the universe of discourse, the determination of lengths of intervals will greatly impact forecasting, effective universe of discourse partition can help improve forecasting accuracy. Based on this, this paper proposes two new methods of universe of discourse partitioning:One method is based on multi-scale ratios to partition the universe of discourse; Another method is based on the adjustable parameter FCM algorithm to partition the universe of discourse. And the two methods are used to forecast Alabama university enrollments respectively, the results show that the new methods outperform the existing some methods.In data fuzzification, a subjective definition method of the fuzzy sets is usually used in present studies, the results make data passivate, cannot reflect data sensitivity, cannot really reflect the distribution of data structure. Based on this, in data fuzzification this article presents a new method of defining fuzzy sets, and gives six new fuzzy sets; Then rules of data fuzzification are given; Finally, through the forecasting of Alabama university enrollments shows that the new definition method is effective.