The Improved SVM Stock Selection Model Based On SMO Algorithm

With the continuous improvement of mathematical finance and the gradual maturity of computer technology in recent years,the quantitative investment in stocks has gradually developed.As a good classification technology,support vector machine(SVM)has been applied to stock selection by scholars,which achieves the classification of stock return and market prediction.However,the data of stock market is huge and complex,so the result of traditional support vector machine is not accurate enough.In this thesis,the sequence minimization algorithm is used to optimize the support vector machine stock selection model,First of all,we select the financial indicators which have a great impact on stock returns,then we use the data that obtained from the dimensionality analysis of these financial indicators to establish a support vector machine stock selection model based on improved SMO algorithm(SMO-SVM).In the construction of the model,the accuracy of SMO-SVM stock selection model in training and testing period is higher than that of SVM stock selection model.Especially for the classification accuracy of high-yield stocks.Then the stock selection model is used to select the stock portfolio.Through the market test of the portfolio,the results show that the performance of the stock portfolio selected by the SVM-SMO model are better than that of the SVM model.Then we divide the market into rising period,fluctuating period and falling period,and the market test of the portfolio is carried out in different period.The results show that in the rising period,SMO portfolio bears high systemic risk and gets high risk return,its return is higher than Shanghai Composite Index and SVM portfolio.In the fluctuating period,SMO portfolio has good stability and basically maintains the excess returns.During the falling period,SMO portfolio has higher expected returns even it bears the same risk with SVM portfolio by the comparison of risk indicesIn this thesis,the algorithm optimization of support vector machine can make it better used in the field of quantitative finance,it can also provide a new idea for the future application of support vector machine.