The Application Of Wavelet Neural Network In Stock Price Prediction

With the quick development of the national economy and the perfection of market economy, stock market became the most important and absolutely necessary part of securities trade and financial market.The Stock market trend's analysis and the forecast have the significant theory significance and the important application value.In recent years,people were generally solve the stock price forecast problem through the setup time series model,time series forecast model in linear system and steady time series question have been widely applied,but in dealing with non-linear features of non-stationary time series on the issue,particularly in some people involved in the initiative system and socio-economic system of forecasting,these methods can not achieve satisfactory results.Along with the development of non-linearized theory and the artificial intelligence technology,the wavelet analysis and the neural network and so on become the money market analysis and forecast tool,but for the wavelet analysis and the neural network intrinsic limitation,they use in the stock price predict that the result is not very ideal,for example,the BP algorithm convergence rate is very slow,if the network initial parameter data selection is not at that time,it is easy to fall into the partial minimum point;The orthogonal wavelet structure is quite complex,can not be easily expressed.But the wavelet neural network unified the wavelet analysis and the neural network's merit,to a certain extent overcame the insufficiency of the both apply respectively in the forecast system's,therefore,the wavelet neural network applied in the stock price forecast that had the important theory significance and the actual application value.In this article, first introduced the basic theory of wavelet analysis and the BP neural network;then use neural network constructed BP neural network short-term forecasting model,the selected model data of the model,the identification of the network input,the methods of data pretreatment and other issues have been Provided,the number of neurons have been implied through the simulation of the model;And do simulation experiment of stocks closing price,the experimental results showed that BP neural network in price forecast is not too bad.Finally embarks from the theory of wavelet neural network's structure,has carried on the analysis of the wavelet and neural network's in loose union way,proposed the method of use wavelet decomposition and the restructuring technology construction wavelet neural network in short-term forecast model's,the criterion has been determined by the simulation experiment of the model restructures;And do experiment to the stock closing price by the simulation,the experimental result indicated that the forecast effect of the wavelet neural network is better than the BP neural network model,has the actual promotion application value.