Font Size: a A A

Research On Stock Price Prediction Based On Investor Sentiment And PCA-GA-BPNN

Posted on:2024-07-17Degree:MasterType:Thesis
Country:ChinaCandidate:X D ZhengFull Text:PDF
GTID:2568307124489494Subject:Financial
Abstract/Summary:PDF Full Text Request
Based on investor sentiment,this paper constructs a PCA-GA-BBN neural network model to conduct stock price prediction research.First of all,in the construction of investor sentiment indicators,this paper takes the text sentiment index of investor text comment,together with the six indicators of consumer confidence index,number of new investor account openings,market turnover rate,trading volume,turnover and price-earnings ratio,as proxy indicators of investor sentiment,and constructs a comprehensive index of investor sentiment after dimensionality reduction processing of data by principal component analysis.Second,this paper optimizes the weights and thresholds of the structure of BP neural network by genetic algorithm(GA),so that the network prediction is easier to approach the global optimal solution,and at the same time improves the generalization ability of the model.Thirdly,this paper explores and constructs the optimal theoretical model model based on the daily data of Shanghai Composite Index and Ling Pharmaceutical as the research object,and comprehensively considers the prediction effect of the model by comparing it with the actual closing price,and uses the four indicators of mean squared error MSE,rmse error,average absolute error MAE and average absolute percentage error MAPE.Finally,through the results of Matlab simulation experiments,this paper proves that the BPNN neural network model(PCA-GA-BPNN)based on PCA principal component analysis and the optimization of GA genetic algorithm is better than the BPNN model,PCA-BPNN model,GA-BPNN model and PCA-GA-BPNN model,which not only has a simpler network structure,but also can reduce risks and improve returns.Through this paper,it is found that adding investor sentiment to the input variable in the PCA-GA-BBN neural network model of stock price prediction can not only improve the accuracy of prediction,but also further reveal the relationship between investor sentiment and stock price.
Keywords/Search Tags:Investor sentiment, Principal component analysis, Genetic algorithm
PDF Full Text Request
Related items