Investigation Of An Integrated Quantitative Selection Strategy Based On Improved Neural Network And Principal Component Analysis

Quantitative investment is one of the important means of investment in the financial field.In the age of big data,The rescue of quantitative stock-picking strategies based on machine learning has become increasingly popular in recent years,Most of the existing stock-picking strategies are based on short-term strategies.In addition,some researches have turned the stock price forecasting problem with large quantity and high noise into a dichotomous problem,and it is difficult to obtain stable excess returns.Based on the improved neural network algorithm in machine learning,this paper proposes A two-stage comprehensive quantitative stock selection strategy and conducts an empirical analysis of the data of China's a-share market in the past two years.In the first stage,GA-BP and Adaboost-BP were introduced as the prediction model,monthly average rise and fall value of stock price was taken as the target value,and multiple factors such as valuation factor and technical factor were taken as the input of the model.The optimal prediction model was selected as the first stage stock selection model through model evaluation.The second stage focuses on the financial data of the company.The financial data of the company corresponding to the stocks selected in the first stage are sorted by the score of principal component analysis.At last,stocks are selected as the final portfolio according to the proportion of stocks selected in the two stages.In order to verify the effectiveness of the proposed method,the prediction effects of Adaboost-BP model,GA-BP model and BP model were compared in this paper.The research shows that the genetic algorithm can effectively find the optimal neural network structure of corresponding data and improve the prediction accuracy of single network structure.The root-mean-square error of the Adaboost-BP model is significantly smaller than that of the other two models,and the accuracy is greatly improved.It indicates that the Adaboost-BP model has better performance in stock price prediction,which is taken as the first stage stock selection model.The effect of applying the comprehensive quantitative stock selection strategy with 80%,100% and 60% of the total stock selection in the stock market is discussed.The empirical results show that the inclusion of principal component analysis in the comprehensive quantitative strategy stock selection can effectively eliminate the companies with poor financial performance.In addition,the selection of appropriate two-stage stock selection proportion strategy has significantly improved the return and stability of the stock.According to the results of comprehensive empirical analysis,the twostage comprehensive quantitative stock selection strategy proposed in this paper can find stocks with more investment value under the consideration of risk factors,so as to achieve the purpose of stable profit and provide investors with good investment orientation.