Prediction Of SSE Composite Index Based On PCA-MLR-MEABP Model

In the development of the stock market,people have been working on how to predict the stock price accurately,meanwhile,forecasting methods are constantly updated.So far,the simple models can not meet people's expectations of the accuracy of the predicting.The PCA-MLR-MEABP combined model proposed in this paper is intended to predict the trend of Shanghai Composite index,and two kinds of integration has been realized: a variety of technical analysis methods are integrated,and linear methods(Principal component analysis,Multiple linear regression)and nonlinear methods(BP neural network which is optimized by Mind evolutionary algorithm)are integrated,which has strong forecasting ability and practical application value.In order to extract the Shanghai stock index trend information,this paper considers these aspects like: the stock price,trading volume,price trend,price range,and chooses eight technical indicators: BBI index,MWVAD index,good rate(BIAS index),K value,D value,BOLL index,WR index,volume.The principal component analysis(PCA)was used to collect and decompose indicators;Then the results of PCA are combined into four principal component variables by multiple linear regression(MLR);Then,the weights and thresholds of BP neural network are optimized by using MEA algorithm,and finally the PCA-MLR-MEABP combined model is constructed.To illustrate the effectiveness of the combined model,four error test statistics are selected from the following two aspects: the fitting precision and the trend prediction precision,and error tests are made between PCA-MLR-MEABP combined model,PCA-MLR model,BPNN model,GRNN model,CV-RF model,and ElmanNN model.Finally,it is found that the PCA-MLR-MEABP combined model has significant advantages in terms of model fitting and trend prediction accuracy.