Research On Quantitative Investment Strategy Of Timing And Stock Selection Based On HMM Model And Multi-Factor Model

Quantitative investment starts late and becomes an important point in the field of investment in recent years.As an iteration of a new generation of research tools,compared with linear algorithms,machine learning has remarkable efficiency,high prediction accuracy and advantages of timely adjusting models according to market changes,which is of great value for investors to improve investment methods and improve investment performance.In this paper,machine learning is introduced into timing strategy and stock selection strategy.Timing strategy obtains Beta returns and reduces risks by grasping the market trend,stock selection strategy obtains Alpha returns by selecting individual stocks,and combines timing strategy and stock selection strategy to control investment risks and improve the stability of returns,providing new ideas for investment practice.First,this paper uses hidden Markov model(HMM)to classify the status of CSI 300 index,and finds that the timing strategy based on this model can effectively avoid bear market,but the judgment of the position is cautious,so it presents the characteristics of strong anti-risk ability and weak profitability.Secondly,this paper selects the multi-factor stock selection model based on asset pricing model in the construction of stock selection strategy,and optimizes and improves the stock selection strategy.While introducing the decision tree algorithm,the traditional multi-factor stock selection model with more advantages in low frequency scenarios is retained,and a more reasonable method is provided for the low frequency factor and high frequency factor to be used together in describing stock returns.That is,a two-level stock pool with monthly adjustment is constructed based on low-frequency fundamental factors,and a portfolio with five-day adjustment is constructed based on medium-high frequency volume price factors.Through the performance comparison analysis,it is found that the portfolio stock selection strategy has stronger profitability than the single model,but the ability to resist risks is poor.When choosing the traditional multi-factor stock selection model,this paper compares the performance of the static equal-weight multi-factor stock selection model and the IC dynamic weighted multi-factor stock selection model.It is found that there is a factor momentum effect in the A-share market,which can create significant positive returns and make the dynamic model relatively outperform.The results show that the portfolio strategy can significantly improve investment returns and reduce risks.Portfolio strategy is significantly better than benchmark index and single timing strategy and stock selection strategy,which can provide reference for individual investors to make decisions.