A Study Of Prediction Technology And Its Application

The stock prediction has important practical significance for government regulationof financial markets and risk prevention of investors in the stock market. The stock marketis affected by the policy, the economy, the psychology of investors, and other complicatedfactors. It is a very complex nonlinear dynamic system with typical features of thecomplex uncertainty. The traditional prediction tools can not meet the requirements ofstock prediction. So people need a stock forecasting method with higher operation rate andprecision. This paper introduces wavelet analysis, artificial neural network model andcombination forecasting method to predict the stock market based on in-depth analysis ofstock market prediction, and for the highly noise and nonlinear characteristics of the stockmarket.This paper uses wavelet transform to pretreat stock time series data. And then thepretreated data is modeled and predicted by combining neural network models and otherprediction models.Firstly, we mainly introduce the related theory of wavelet analysis, artificial neuralnetwork and combinatorial forecasting model and research maximal overlap discretewavelet transform, maximal overlap discrete wavelet packet transform and the basic ideaof combinatorial forecasting model. On this basis, we give two improved waveletthreshold denoising methods and an improved wavelet neural network model, and twoexamples are used to verify that these improved wavelet denoising methods and theimproved wavelet neural network model are effective and superior.Secondly, we introduce variance reciprocal method and optimal variable weightcombination forecasting method. On this basis, we propose three new combinationforecasting methods: new variance reciprocal method, SMNE stationary weightcombination forecasting method and the error reciprocal variable weight combinationforecasting method based on artificial neural network. Finally an example demonstratesthat the proposed methods are feasible and effective.Furthermore, based on wavelet analysis, wavelet denoising, neural network models and combination method, this paper gives several kinds of combination forecastingmethods based on wavelet transform, and details the prediction process of combinationmethods through flow charts.Finally, we use the proposed several prediction methods to predict Shanghai andShenzhen stock series. Compared with the traditional time series analysis method or asingle model to predict stock, the stock forecasting effect has been greatly improved.