Algorithm Research On Stock Selection And Timing Quantitative Investment Strategy Based On Machine Learning Algorithm

Quantitative trading has been developed for over forty years in developed countries abroad.In recent years,with the advancement of computer technology,quantitative investment has gradually become one of the popular trading methods in China.It has become increasingly mature,and how to find a more stable and profitable trading strategy has always been a topic of great interest for scholars and investors at home and abroad.An excellent trading strategy can accurately allocate investors’ assets,seize favorable investment opportunities,and keenly judge the overall trend of the macro financial market,thereby obtaining higher returns.Stock selection and timing are two important strategies in the field of quantitative investment.Among them,stock selection strategy refers to building a suitable investment portfolio for profit,while timing strategy refers to selecting the appropriate buying and selling time points for profit,and organically combining the two strategies from different perspectives.Not only can stable returns be achieved,but investment risks can be controlled,and the operation of the stock market can be stabilized to a certain extent.Quantitative trading cannot be separated from the collection and processing of big data.Machine learning algorithms have natural advantages in processing big data.Therefore,this article attempts to construct quantitative strategies and combination strategies for stock selection and timing based on machine learning algorithms.This article elaborates on the principles of stock selection strategy and timing strategy,with the Shanghai and Shenzhen 300 component stocks as the research subjects.The sample study period is from January 1,2018 to January 1,2023,which covers bull,bear,and volatile markets.The stock selection strategy is constructed based on the factor database data of the ricequant quantitative investment platform.Through the single factor test and correlation analysis of the factors,the selected financial and technical factors are used to construct the stock pool of the stock selection strategy.The decision tree algorithm,random forest algorithm and KNN algorithm are used to implement the factor stock selection strategy at the daily frequency level and analyze the performance of the stock selection strategy backtesting;The timing strategy part is constructed based on data from the MindGo platform in Tonghuashun.The data interface of Kaiyuan is called to obtain and calculate the corresponding eigenvalues,which are used as observation vectors for the hidden Markov model and undergo parameter training to identify the predicted index state and make buy or sell operations.From the analysis results,it can be seen that the selected investment target can obtain excess returns when the risk is controllable,And the relevant volatility risk indicators are relatively stable;Finally,a combination strategy was constructed by combining timing strategy and stock selection strategy,and the performance results of the combination strategy were compared with stock selection strategy and timing strategy.It was found that the combination strategy was superior to a single strategy in terms of both revenue and risk control.This article implements stock selection strategy and timing strategy through machine learning algorithms,and verifies the feasibility and effectiveness of the combination strategy of the two,providing investors with new investment ideas.