A Research On Stock Price Prediction Based On PCA-GA-BPNN

Since the birth of the stock market,many scholars at home and abroad have devoted themselves to studying the stock market,and have put forward many methods for predicting the corresponding stock prices.However,the stock price will be affected by many factors such as the macroeconomic environment,related industry conditions and company operating conditions.And there are many complex nonlinear relationships between the factors that affect stock price volatility.The variability of these internal factors and external factors has led to a very arduous job of stock price forecasting.This article is to analyze the various stock price forecasting methods and combine artificial intelligence with this hot topic.The BP neural network method is used to forecast the stock price and realize it through the matlab program.The BP neural network has self-learning and non-linear approximation capabilities,and this just can solve the nonlinear relationship between the stock price factors.The BP neural network can learn the historical data of the stock,so as to find out the inherent operating rules of the stock market and predict the future trend of the stock price.However,there are still some deficiencies in itself,such as the problem that there may be data redundancy in the input data,it is difficult to find the global optimal value,it is often a local optimum,and there are also problems such as slow training rate.Therefore,this paper applies an optimization method to the defects of BP neural network to improve its own defects.Firstly,the input sample data is reduced by using the principal component analysis method.On the one hand,the redundancy between data can be avoided.On the other hand,the dimension of data input can be reduced and the operation efficiency of the algorithm can be improved.Then it is aimed at BP neural network.The problem of subsidence is locally optimal.The GA algorithm is used to optimize the weights and thresholds of the network.Finally,the sample data selected in this paper is the CSI 300 stock index daily data,which spans from April 16,2010 to September 7,2017.Through the matlab simulation experiment,the topological structure of the BP neural network and the number of nodes in the hidden layer are determined and the corresponding parameters are determined.The stock price is predicted and the mean square error and relative error percentage of the predicted,actual and predicted values of the network are output.The result of matlab simulation can prove that the prediction accuracy of neural network optimized by principal component analysis and genetic algorithm is more accurate,and the model is more stable.This model is feasible and effective for stock price forecasting.Especially for small and medium-sized investors who invest in stocks in the stock market,they can be used as an effective stock price forecasting method to assist them in investing,as far as possible to reduce their risks and increase their earnings.