Improvement Of Fuzzy Time Series Forecasting Model And Its Application

Fuzzy time series forecasting model,which was proposed to solve the fuzzy problems that the classical time series analysis methods cannot deal with,is an extensive research topic on data analysis and prediction.The continuous development of fuzzy set theory has promoted more and more in-depth ap-plication of fuzzy time series forecasting model.The progress of Information Age makes it more and more widely used,and how to improve the prediction accuracy of models is attracting the attention of numerous researchers.Based on fuzzy set theory and time series analysis theory,the paper studies fuzzy clustering,fuzzyization,establishment of multivariable fuzzy logic relationship group,de-fuzzification and other aspects of fuzzy time series forecasting mod-els.The main contents are as follows:1.In view of fuzzy c-means clustering algorithm(FCM),in order to avoid falling into the local minimum and overcome the sensitive barriers to the initial clustering center,an improved FCM is proposed.Revise the initial clustering center,type-2 fuzzify time series and establish the fuzzy relation matrix,then the final prediction value is obtained by weighting the clustering center and predicting the first difference of time series.Finally,this method improves the prediction accuracy and shows its effectiveness and superiority by comparing the prediction of Alabama university enrollments and Taiwan Stock Exchange Capitalization Weighted Stock Index(TAIEX)forecasting model.2.Considering that many time series models in real life are affected by multiple variables,and some are even influential,multivariate time series fore-casting model is further studied in the article.Because the geometric structure and fuzzy partition of different time series may be different,this paper uses fuzzy clustering validity index based on Mahalanobis distance and obtains the optimal clustering number.By constructing the first-order multivariable fuzzy logical relationship group and de-fuzzifying the predicted fuzzy value to the actual value of the previous moment and related clustering centers,the final prediction value is obtained.Finally,take TAIEX as an example,the new prediction method is compared with other prediction methods.The results show that not only the prediction error of new prediction method is smaller,but also the stability is better.