A Study Of Portfolio Selection Based On Stock Forecasting

Stock return forecasting and portfolio selection are important branches of stock market research.An effective stock return forecasting model can help investors to clearly understand the future return trend of stocks,while a better portfolio selection model can help investors to gain more excess return with the same level of risk.Therefore,it is a very meaningful research to combine stock return prediction with portfolio.Currently,some scholars have constructed portfolio selection methods based on machine learning and deep learning methods and achieved better results.However,most of the existing studies on stock forecasting use a single model and do not effectively utilize the useful information implied in the forecast errors.Therefore,in this paper,a portfolio selection study based on stock return forecasting is conducted using a combination forecasting approach.The main research works are as follows.(1)A combined forecasting model based on three machine learning methods is constructed.First,a combined prediction model based on support vector machine(SVM),random forest(RF)and gated recurrent unit neural network(GRU)is constructed to predict stock future returns.Secondly,an equal-weight combination method is used to combine the predicted values of each model to obtain the predicted values of stock returns,and then the high-quality stocks are screened out.Finally,the effectiveness of the combined forecasting model proposed in this paper is verified by empirical analysis.(2)An improved portfolio selection model based on return prediction error is proposed.First,metrics for measuring risk and excess return are proposed based on the forecast error.Further,an improved multi-objective portfolio selection model based on the mean-variance model is proposed,and the constructed multi-objective model is transformed into a single-objective model.Finally,the effectiveness of the improved portfolio selection model proposed in this paper is verified by empirical analysis.