The Research On Forecasting Of Time Series Based On Wavelet Analysis And Optimization Theory

Time series forecast is one of the current forecast methods and it is an applied research field with some practical value. The problem of predicting the future data on the basis of historical data exists in many practical applications, especially in economy, engineering and so on. Therefore much attention has been paid to it. In many practical predictions, time-series are often highly complex and presented nonlinear state. For that reason, the predictive result of single prediction method is not satisfied. In order to achieve better forecasting results, people attempt to combine different models. In this paper, the prediction model is combined ARIMA model and Elman neural network. Moreover, both the conjugate gradient methods and wavelet analysis theory are used to improve the single model. The paper checks the combination prediction model by an example and the results of the example validations indicate that the method mentioned in this paper is better than single model.The wavelet denoising is applied to preprocess the history time series data of the stock. Then, the preprocessed series are modeled and forecasted by combining time series models and neural network models.First, the paper introduces the related theories of time series models, nonlinear conjugate gradient method, wavelet analysis and neural network. Meanwhile, the paper studies the nonlinear conjugate gradient method and proposes an improved conjugate gradient algorithm. Then on that basis, the improved conjugate gradient algorithm is successfully used to parameter estimation of time series models so as to optimize the parameters of the model. The programming effectiveness and feasibility of the method is verified through an example.And then the theories of the wavelet threshold de-noising are studied. Through the research, the paper proposes a new denoising method. By the end of the chapter, examples are given to prove the efficiency and superiority of this method.Furthermore, the theories of wavelet neural network is been introduced and a new combined forecasting method based on wavelet neural network and time series model is presented in the paper. Moreover, the detailed forecast process is presented through the flow chart.Finally, the proposed prediction method is applied to predict stock series. Compared with a single model, the results show that the method proposed in the paper has better prediction effect.