Prediction Of Stock Price Trend Based On BP Neural Network

This thesis analyzes in detail the principle and learning algorithm of BP neural network,applying three layers of BP neural network to the Shanghai Composite Index in the empirical analysis.Firstly,by examining the rank correlation coefficient between the influence factors and the Shanghai Composite Index closing price,factors are screened in order to input more effective factors,get more efficient mapping,and discover the inherent law between the input and output;Secondly,compared with the standard BP neural network and the modified BP neural network for the Shanghai Composite Index closing price of empirical analysis,through the average absolute error(MAE)and direction accuracy(DS)between the predicted values and the real value,concluded that the modified BP neural network in the prediction ability better than the standard BP neural network;Then,this thesis using the modified BP neural network through the mean absolute error(MAE),direction correct(DS)to determine the best parameters,namely the training set,training function,learning rate,the number of hidden nodes and activation function;Finally,this thesis compares the prediction ability of BP neural network in volatile and bull market,from which we observe that the performance of the predict in the volatile maket is not as good as the bull market.Based on the BP neural network theory and empirical research,we find that the three-layer BP neural network can achieve effectively prediction of the stock price and has certain reference significance to the investment practice.