Studies Of Nonstationary Time Series Prediction Method Based On Wavelet Analysis

Predicting is the process of studying historical events, building system models onprevious results and making predictions. With the developing of science and technology,the prediction methods have been greatly improved. However, in the practice of prediction,fitting predictions can not be conducted well by single prediction method. Based on thissituation, in order to improve the prediction accuracy of the non-stationary time seriesmodel, the predicted methods were studied in this paper. This paper introduces waveletanalysis theory, neural network theory and combination forecasting methods to analysisand predict non-stationary time series. The modification method were put forward in thispaper is used in our tertiary industry analysis. The combination forecasting models basedon wavelet analysis and neural network prediction are used in different areas such asQinhuangdao tourism population. In the course of study combination forecasting modelsfor solving the variable weighting coefficient, the time weight were adequately taken intoaccount. The actual data fitting results prove the feasibility and effectiveness of thismethod.Firstly, this paper introduces the theory of time series analysis model, starts from theidea of combined forecasting and cites the metabolism of GM (1,1) model in gray theoryto modify the prediction results of the ARIMA model. It improved the predictive accuracyof the model.Secondly, it introduces the theory of wavelet analysis and wavelet de-noising.Wavelet thresholding algorithm is studied. In order to illustrate the wavelet de-noising hasgood features that can be used to non-stationary time series de-noising. By example, thismethod is feasible.Furthermore, it introduces the relevant knowledge of neural network model. Thewavelet decomposition and neural network are formed a combination of wavelet neuralnetwork combined forecasting model to analysis and forecast Qinhuangdao touristpopulation time series. Compared with the traditional time series analysis method or asingle prediction model, the prediction results have been greatly improved.Finally, this paper introduces the basic idea of combination forecasting model. Basing on variable weight combination forecasting method, it proposes a time weightedcombination forecasting method and variable weights were solved. By the end of thechapter, examples prove the effectiveness and feasibility of the method.