Research On Stock Price Trend Forecasting Method Based On Deep Learning

Today,stock market is not only providing financing for good listed companies,but also providing a way out for some investors with a sense of investment.So that social resources are better configured and macro-economic are controlled.However,because of the uncertainty of the stock market and every investor the cognitive vision same-sex and technical analysis of the complexity of the factors,makes the general investors investment returns short of expectations on the stock market,some even lose everything.So the stock market has been heavily focused on government,business,and investors.The forecast of stock price trends is a hot spot in stock research.As is known to all,due to the volatility of the stock market has many characteristics,such as strong nonlinear,high noise,so is extremely difficult to stock price trend forecast,the traditional stock forecasting methods often with little success.So how to set up new stock price trend forecast model to enhance the accuracy of prediction,thereby helping to effectively avoid risks,financial investors investment profit maximization,has important theoretical significance and application value.This article first expounds the traditional stock prediction methods.The basic analysis mainly from the macro micro economy,the related information such as the company's financial statements and cash flow Angle,through the relative valuations and discount valuation,estimates the intrinsic value of the shares.Disadvantages: the information is not accurate,such as the delay,accuracy of the information.The market analysis is mainly on the basis of statistical charts,such as K line graph,its shape can be divided into the form and trend line,according to the specific form to determine the future trend of the stock market.Disadvantages: there is a wide variety of methods of analysis,and there is a huge difference between the methods.Statistical analysis method is mainly the least squares was used to construct various regression,such as mixed regression model,autoregressive model of stock price trend forecast,this kind of model prediction accuracy compared with the previous two kinds of forecasting methods.Disadvantages: the regression model is usually too many assumptions,and problem for strong nonlinear processing ability,and the stock price trend forecast problem affected by many factors and strong nonlinear.Prediction model based on artificial neural network has high self organizing,self adjusting and self learning ability,and is a complexity of extremely high nonlinear system,model predicted results is usually better than the traditional method.Disadvantages: stock prediction model based on neural network easy to fall into local minimum problem,and multi-layer neural network in the description of the complicated things,tend to increase the layer number of hidden layer,this will lead to gradient diffusion problems,which influence the accuracy.In this paper,prediction model restricted Boltzmann machine is used to build a deep belief networks,learning method is to use K step after Gibbs sampling,combined with the contrast divergence algorithm,to train the deep belief networks.Finally,this paper USES the stock price information of Gree air conditioner to train the prediction model and test the prediction accuracy of this model.Selection based on BP neural network predictive model of stock price prediction model as this paper contrast model,by using examples of Guizhou Maotai and Byd's stock price information to verify the prediction exactness rate of the two models,the experimental results show that the stock price trend prediction model based on deep learning effect is good,and the accuracy is better than BP neural network prediction model.The innovation of this paper:(1)this paper adopts the Boltzmann institutions built based on the limited depth prediction model of stock price trends belief network,study methods adopted after K steps of Gibbs sampling contrast divergence algorithm(CD)algorithm to train prediction model.Finally,the example is verified.(2)this paper compares the prediction accuracy of prediction model based on BP neural network stock price trend prediction model.