Research On The Fuzzy Time Series Forecasting Model Based On Improved Artificial Fish Swarm Algorithm

Now we are in an era of information explosion,that how to use the collected data to predict the behavior trend of the future development of things has become a research hotspot in the field of data analysis.In the real world,lots of data is incomplete,uncertain and fuzzy etc.Fuzzy time series prediction model(FTS)demonstrates its unique advantages when dealing with these data,which attract more and more researchers devoting to the research and application of the model.At present,the fuzzy time series prediction model has been widely used in temperature prediction,traffic flow prediction,electric power prediction,network user prediction,tourist quantity prediction and other aspects.After a large number of studies,it's found that the domain partition and fuzzy prediction method are two important factors to influence the prediction accuracy of the model,therefore,this article starts to improve the fuzzy time series prediction model from these two aspects.In terms of domain partition,propose the hybrid artificial fish swarm algorithm(HAFSA)based on bacterial foraging algorithm(BFA)and Levy flight(Levy),and use them for the determination of interval lengths.In order to verify the effectiveness of the proposed algorithm,this paper adopts three normative test functions to simulate and compare the optimal performance of HAFSA algorithm and the classical artificial fish swarm algorithm(AFSA),the results show that the prediction accuracy and the convergence speed of HAFSA algorithm improve obviously.In terms of evaluating the predicted value,as the observed value was the key factor to determine the predicted value in the past,so how to effectively collect the past observed value,has become the research focus of the fuzzy time series model,this paper proposes to utilize ordered weighted averaging operator(OWA)to collect the past observed value,and puts forward a fuzzy prediction method based on ordered weighted averaging operator,and makes a comparison of the fuzzy prediction method(MV)raised by Kuo etc,further illustrates the effectiveness of the fuzzy prediction method.Finally,in order to verify the validity of the improved fuzzy time series prediction model proposed by this paper,take registered quantity in Alabama University as the sample data to conduct the simulation experiment,the results show that the proposed algorithm has higher prediction accuracy.