A Research And Application On Discourse Of Fuzzy Time Series Models

In the era information explosion, there are increasing number of means to obtain data, but the credibility and accuracy of data is on decreases. Owing to the features which have a good robust for uncertainty and ambiguity historical data of fuzzy time series, forecasters are sucked by it.With the increasing widely application of fuzzy time series models, there are more and more researchers dedicated fuzzy time series. According to the prediction step models, there are two research direction, one is the division of discourse, the other one is fuzzy logic relationship.On the definition of discourse, this thesis studied the method of discourse intervals. Due to the existing research on the definition of discourse is only simply rounded up and down, unaware of the importance of the definition of discourse can also affect the result of prediction, so this thesis puts forward a new method about the definition of discourse. In this new method, the definition of discourse is related to the data distribution of current category. Because the distribution of the sample data is considered, so the accuracy of intervals is improved.This thesis makes the improvement based on multi-scale ratio discourse partition and puts forward using adjacent data geometry mean instead of arithmetic mean to calculate the ratio. Although the method of multi-scale ratio discourse partition made full use of distribution of the data by using FCM algorithm, there is still a situation uneven distribution of an inner class. At this time, the advantage of the geometric mean is that it is used to calculate the ratio, so using adjacent data geometry mean instead of arithmetic mean to calculate the ratio effectively improve the accuracy of the classification of classification of the interval, improving the accuracy of prediction.In order to prove the effectiveness of the proposed method, this thesis predicts the enrollment Alabama and traffic, and the result of experiments show that this method has good prediction effect.