The Application Of Dynamic Weighted Multi-Factor Model Based On Cluster Analysis

The multi factor model has undergone decades of development since its birth and is still the mainstream quantitative stock selection model in the market.CSI300 futures have been traded since April 16,2010.Although there are still many restrictions in futures trading,such as margin ratio and high transaction fee,it is possible to obtain excess returns through factor selection.In recent years,due to the support of academic theory,the changes in the trading environment,the development of computer technology,the updating of algorithms and the explosive growth of the scale of data,the artificial intelligence technology represented by machine learning algorithms has gained more attention,including the use of neural networks,Support Vector Machines,cluster analysis,Hidden Markov models and other machine learning models are used for data analysis and stock selection.The application of machine learning model will enable us to discover more complex data structure which has not been discovered before and improve the efficiency of data utilization.The first chapter is the introduction.It introduces the development of quantitative investment and multi factor models,as well as the main contents,structure arrangements and innovations of this paper.The second chapter mainly introduces the research status domestic and abroad,analyzes the theory of the multi factor selection model and cluster analysis,and lays a theoretical foundation for the follow-up empirical research.The third chapter carries out single factor test on 97 factors,such as valuation,growth,revenue,quality,technical,and so on from three aspects of factor forecasting ability,factor revenue and factor monotonicity.Select 69 effective factors and build on the traditional multi factor selection method in the China stock market.The result of the portfolio is a benchmark for comparison of subsequent portfolios.The fourth chapter,cluster factors based on the correlation of factor information connectedness and construct a portfolio.Calculate factor information ratio and build another portfolio weighted by factor information ratio.Compare these portfolios.The results of the four stock selection methods show that,under the same factor classification,the revenue of the portfolio weighted by factor information ratio is obviously higher than that of the equal weight portfolio.Under the same weighting method,the revenue of the portfolio classified by clustering analysis is higher than that of the traditional portfolio.The clustering analysis of factors can track market style changes through the change of the factor information connectedness.Based on factor information ratio,the factor weighting can be used to identify the share of the factor,and then increase the portfolio revenue.Based on the multi factor model,this paper combines clustering analysis and factor weighting to improve the utilization ratio of data and optimize the portfolio performance.The application of machine learning model is a trend of quantitative investment in the future.As a common unsupervised machine learning model,clustering analysis is introduced into multi factor selection model for the combination of machine learning and multi factor model,which has some reference significance.