An Empirical Study On The Applicability Of Improved Fama-French Model Based On Machine Learning In CSI 300 Market

Asset portfolio pricing has always been an important research direction in the financial field.This paper explains the impact of asset return from the perspective of risk and return,and studies the relevant factors that determine the return on assets.Capital asset pricing model is based on strict theoretical assumptions,and the stock markets of different countries have different degrees of satisfaction with the assumptions,so there are significant differences in applicability.Machine learning is a comprehensive science and technology.It simulates human's behavior and thinking mode based on the method of statistical probability through the fast calculation speed of computer,and automatically learns new knowledge from it,so as to improve the performance continuously.As the latest technology in machine learning,deep learning has provided a new driving force for the research of quantitative investment.This paper selects Fama French five factor capital asset pricing model as the research object and modifies it by machine learning.Through quantitative analysis,autoregressive model and business cycle model,it is determined that the main sample is CSI 300 component stocks,the sample frequency is monthly data,and the sample time span is April 2005 to April2018.The applicability of the five factor model in CSI 300 market and the explanation degree of the five factors are studied from the perspective of individual stock and time cross section.It is found that there are two redundant factors in the model,so we delete them and construct a three factor model.Finally,a new factor is selected by machine learning model support vector machine,random forest and logical regression,and added to the three factor model to construct the final four factor model.The results show that the model has a higher degree of fitting,the explanatory power to a certain extent,can be better applied in the CSI 300 market.At the same time,due to the characteristics of CSI 300 market itself,it is concluded that the market has scale effect,value effect,profit effect,investment effect and transaction effect.In view of the results of theoretical research and empirical analysis,this paper puts forward some countermeasures and suggestions to standardize asset pricing and management in China's securities market: strengthening investment education to avoid blindly following the trend;speeding up financial reform and increasing investment tools;improving financial control and improving supervision level.