Research On Stock Volatility Prediction Method Based On Wavelet Analysis And BP Neural Network

Stock price has natural randomness and contingency.The volatility of stock price usually appears as an irregular random fluctuation.With the addition of artificial psychological factors and uncertainties in the entire social environment,we can see a single linear time series prediction models,such as Exponential Smoothing Method,ARMA Model,and Moving Average Method,could not achieve satisfactory prediction effect in stock price forecasting because of the limitations of the stability,normality,and independence of the time series.Therefore,the introduction of nonlinear analysis method and the integration of various models in stock price forecasting have become a new inevitable choice in the research field of stock price forecasting.BP neural network has a unique advantage in stock forecasting.It can be used for prediction without establishing explicit relationships and model of complex nonlinear systems.It is a relatively advanced theory in the current prediction theory.However,researchers soon found that BP neural network model is easy to have local minimum values,slow convergence of the algorithm,difficulty in explicit expression,and other problems if selecting inappropriate initial parameters of the network.As a result,the prediction accuracy rate is reduced,and even the result of the forecasting trend is wrong.These serious errors have affected the further use of neural network model in stock forecasting.Wavelet analysis has been widely used in various fields of science and engineering in recent years.It can be proved in practice that wavelet analysis has more obvious advantages than traditional noise reduction methods in terms of data noise reduction because it can localize time-frequency.Therefore,it is also an inevitable choice to introduce wavelet analysis into the economic and financial fields as an analysis and forecasting tool for stock market.However,although wavelet analysis has a wide range of applications,the use of wavelet analysis alone has rarely achieved excellent results in practical applicationBecause wavelet analysis and BP neural network have their own inherent limitations,the blind use of any kind of analysis method may lead to the wrong prediction of the stock price,and causing financial losses to investors.Therefore,starting from the theory of wavelet neural network construction,a "short-term prediction model of wavelet neural network" is formed by combining wavelet decomposition and reconstruction with BP neural network.By preprocessing and processing the collected data,this model performs short-term analysis and trend prediction of the closing prices of individual stocks.And it turns out that the wavelet neural network prediction model is effective and feasible.It is hoped that the research of this paper has a certain theoretical and practical significance to the analysis of stock investment strategy.