Improved Quantitative Stock Selection Model Based On PCA-SVM

With the continuous development of China's securities market,more and more institutional investors and ordinary investors have entered the market.In the face of a variety of complex investment products,how to choose good products is an importan-tissue.The two major analytical methods that professional investors are very much interested in are fundamental analysis and technical analysis.Nowadays,the rapid development of computer technology has promoted the continuous progress in the field of quantitative investment.More and more investors are beginning to assist in stock selection by mining a large amount of data and establishing a quantitative mod-el of the system.The huge advantage of quantitative investment is increasingly clear.In this paper,we use a composite algorithm combining principal component analysis and support vector machine to construct a quantitative model.In the empirical resea-rch,firstly,the feature extraction of 19 stock indexes is carried out by the good dime-nsionality reduction ability of principal component analysis,and the extracted result is used as the input set of support vector machine to construct the support vector mac-hine stock selection model.Tests were performed on the test set to examine the gene-ralization capabilities of the model.In the process of optimizing the SVM model par-ameters,the empirical research uses a combination of genetic algorithm and cross-validation method.Finally,a single SVM classifier model is constructed as a comp-arison model.The results show that the classification accuracy of the constructed PCA-SVM model is as high as 86.85%,which is much higher than the single SVM model.The empirical study selected the weights of 5% of stocks classified as high quality stocks to construct a portfolio,and compared it with the Shanghai and Shen zhen300 Index(benchmark)from the perspective of investment yield,and found that the selected stocks have better profitability.