Research Of GA-based Generalized Fuzzy Time Series Modeling With Its Application

A GA-based generalized fuzzy time series mode is proposed in the dissertation, realizing a division of the universe of discourse with 7 fuzzy sets. By these membership functions of fuzzy sets, sample data are fuzzied. Then, we obtain the various fuzzy states and fuzzy logic relationships of different levels. For calculating conveniently and simulating the law of recognization vividly, the model just consider several main fuzzy logic relationships of difference levels. Subsequently, using the predicting for tourism requirement of Huangshan, the proposed model is verified by the empirical analysis.Firstly, a summary of the research background and current situation of fuzzy time series are given. This dissertation expound some common fuzzy time series model and give the relevant concepts and definitions.Secondly, in the third and fourth chapters, the calculation of the first-order differences of the number of tourists is executed firstly. Then, the first-order differences of the number of tourists will be treat as a new time series for study, and fuzzied by triangular fuzzy numbers and Trapezoidal fuzzy number respectively. Subsequently, a forecast is used for tourism requirement of Huangshan by the proposed model in this dissertation. Compared with the traditional fuzzy time series models or the hybrid model, the proposed model can overcome the subjective randomness of partition interval and improve the accuracy of prediction. Moreover, the proposed model present a better forecast performance when the time series of Huangshan tourists fuzzied by Trapezoidal fuzzy number.Finally, the main content of fifth chapter is the prediction of interval number time series. First, transfer Huangshan tourist number time series to interval number time series. Second, calculate first-order differences of interval number time series by interval arithmetic. Then, 7 Trapezoidal fuzzy numbers are defined on the universe of discourse and applied to calculate the membership degrees of the interval number time series. Executing the predicting for these interval number time series, the propose model also present a better fore performance.