Stock Timing Research Based On BP Artificial Neural Network Model

Stock timing is important for financial market investors and is the key to success.It is the focus of research by experts and scholars.For a good stock,if do not accurately grasp the chance of timing,you will also face investment risks.Stock timing is very complicated,because although the price trend is not completely random,it is affected by many factors,so it is difficult to judge accurately.The traditional investment method can only be analyzed from some basic information and simple technical indicators such as KDL and moving average,and the accuracy is relatively low.Therefore,the more popular investment method in the West is quantitative investment.Quantitative investment is an investment method that combines financial investment with mathematical models,mines information on big data,and then uses computers to automate transactions.It makes transactions more systematic,accurate,efficient,and objective,and can improve the efficiency and accuracy of investment decisions.Among these models,the neural network model has a strong nonlinear approximation function,which has great advantages for dealing with nonlinear and complex stock markets.Therefore,this paper first expounds the relevant theory of neural network model,and then establishes the empirical analysis model of neural network,the specific content is as follows.This paper takes the Shanghai Composite Index as the research object,and selects the daily data of its 16 commonly used stock indicators.The time interval is from January 5,2015 to September 7,2018,with a total of 900 data sets.By establishing a BP neural network model for the original data,predicts the daily closing price of the Shanghai Composite Index.In order to make the use of data more simple and accurate,the principal component dimensionality reduction analysis is performed on many original variables.The raw data was tested using the KMO test and the Bartlett spherical test to verify its suitability for principal component analysis.After the test,the raw data was subjected to principal component analysis.According to the results of principal component analysis,six principal components are selected,that means these six principal component data can represent most of the information of 16 stock indicators.After the variables are dimensioned,establishes the BP neural network model.The input variable of the BP neural network model is 6 principal component data,and the output variable is the daily closing price.The number of hidden layer nodes is determined to be 5 according to the hidden layer node formula.After building the model,start making predictions.The prediction is then compared to the actual data and its relative error is calculated.In addition,in order to compare the accuracy of the model,the ARMA-GARCH model is established for comparative analysis.The analysis results show that the prediction error of ARMA-GARCH model is higher than that of BP neural network model based on principal component analysis,and the prediction accuracy is relatively poor.Therefore,the BP neural network model based on principal component analysis has reference value for stock market timing.