Time Series Forecasting Model Based On Fuzzy Neural Network

Time series forecasting is a subject of technology,which aimsto forecast and control through building model by some limited historyobserved samples and use it to explain the law of the data,and have awide of applications in many fields.For stationary time series model-ing and forecasting,particularly the research of the linear model,manymature technologies and methods have been found.However,in practi-cal problems,majority of time series are not stationary and linear,andthere is not good method to analysis and deal with such time series andto less than the desired effect.At the same time, the fuzzy neural net-work model which has good performance in non-linear forecasting andparticularity has very good achievement in tourism,financial,industryand other ?ieds.But in reality,many time series are not simple lin-ear or non-linear models,and it is difficult to judge in actual opera-tion,furthermore there is not a single model suitable for any situa-tion.Therefore, the idea of hybrid model have been proposed by somescholars such as forecasting model using hybrid fuzzy neural networkand time series, forecasting model using hybrid fuzzy neural networkand time series, and have achieved good results.A time series forecasting model based on fuzzy neural network havebeen proposed in this paper.the development of time series and fuzzyneural network have been introduced in the introduction part.In chap-ter 2,the basic method of traditional time series analysis is given suchas AR model,MA model and ARMA model.For non-stationary lineartime series,they can be converted into stationary sequences through mathematical methods such as di?erential,such as SARIMA model.Forthe financial time series,ARCH model is given.In chapter 3,the fuzzyneural network model,its system structure and its learning algorithmare given,it have a good ability to predict non-linear time series.Thelast two chapter are important part of this paper,in chapter 4,a time se-ries forecasting model using a hybrid fuzzy neural network and SARIMAis given,and the specific hybrid model is given according to the avia-tion data in Sydney.In chapter 5,a time series forecasting model usinga hybrid fuzzy neural network and GARCH is proposed,and the spe-cific hybrid model is also given according to the Shanghai Stock Indexdata.According to the experiment data, the hybrid models are used andcompared among four other models, i.e., the time series model,theFNN model,the Support Vector Machines model and the neural net-work model.The experiment result indicates that the proposed modelis effective in the parameter choice,algorithm stability and precisionforecasting.