A Study On Quantitative Investment Strategy Based On Penalty Constraint Method

With the rapid development of the financial and computer technology,quantitative investment has become more and more concerned,and has become a hot issue in the securities industry.The stock market is a complex nonlinear system.The traditional algorithm is used as a tool for stock price forecasting and investment guidance.It has great limitations,and the traditional quantitative model is single.The machine learning algorithm is used to construct a quantitative investment model for stock price forecasting.This paper uses the stocks of the HS300 as stock pools.The factor cross-sectional data of the last trading day of each quarter from January 2013 to December 2017 is used as a sample of all data.A quantitative investment strategy based on the penalty constraint method is constructed and the historical backtest results of the strategy are evaluated and analyzed,and a risk control optimization scheme is proposed to obtain a robust excess return.Firstly,this paper selects the factor data at the end of the first quarter of 2017 and the yield data of the HS300 constituent stocks in the second quarter of 2017 as sample data,which selects the quality factor,the value factor,the emotion factor,and the technical indicators.A total of 42 factors of factor and growth factor are used as input variables,and the future yield of stocks is used as an output variable to construct a regression prediction model.The multiple linear regression model and the LASSO regression model based on penalty constraint method and elastic network regression are compared.The prediction accuracy of the method proves that the prediction model based on penalty constraint method is better than the traditional multiple linear regression model,and the elastic network regression is superior to the LASSO regression model in model prediction accuracy.At the same time,the idea based on penalty constraint combines with the main Component analysis and LASSO regression,the improved regression prediction method PCA-LASSO model is proposed and used to predict the future earnings of stocks.In theory,thePCA-LASSO model not only eliminates highly correlated,but also achieves the goal of dimensionality reduction.The empirical results also show that the PCA-LASSO model is optimal in model prediction accuracy.Then,based on the above four models,the paper uses the rolling prediction method to construct a multi-factor quantitative stock selection model based on regression method.According to the forecast results of each model,the top 50 stocks will be used as the stock portfolio of investors in the next period,and the historical backtest will be tested.The entire backtest time is from January 31,2013 to 2017.On the 31 st of the month,the strategy is to adjust positions on a quarterly basis.The backtest results show that the annualized rate of return based on the PCA-LASSO method is 45.4%,the annualized rate of return based on elastic network regression is43.2%,and the annualized rate of return based on LASSO regression is 40.4%,and the benchmark annualized The yield is 15.7%,which indicates that the multi-factor stock selection model based on the penalty constraint method outperformed the HS300 Index,and the improved PCA-LASSO model also performed best during the historical backtest.Finally,based on the multi-factor quantitative stock selection,quantitative timing strategy is designed.The empirical research shows that the annualized rate of return based on the double-average trading strategy is 28.8%,and the annualized return rate of the benchmark income is 15.7%.It can be seen that the quantitative timing income based on the double-average technical indicators has outperformed the HS300 Index.The maximum basement of the double-average strategy’s return is6.3%,while the maximum basement of the HS300 index is 108.8%.Obviously,the risk of the moving average strategy is smaller than that of the HS300 index.It can be seen that when the stock market volatility is relatively large and relatively frequent,the quantitative timing strategy helps investors obtain stable positive returns while also controlling risks more effectively.