Multi-factor Quantitative Model Optimization Based On Factor Validity And Fitting Method

In recent years,with the maturity of my country's capital market,the number of rational investors has increased,and the demand and application for quantitative stock selection have become more and more extensive.Among the many quantitative stock picking theories,multi-factor stock picking strategies are favored by institutional investors due to their inherent characteristics such as discipline,system,and accuracy.However,with the rapid switching of market styles,some factors have failed in recent years.At the same time,stock selection models based on these factors have also experienced phenomena such as poor yields and large drawdowns.This paper establishes a set of targeted multi-factor stock selection models based on such problems.This article selects the Shanghai and Shenzhen A-share stocks as the research object,and uses the monthly data from 2015 to 2019 as a reference basis for factor validity testing.First of all,this article selects seven categories of style factors from the fundamentals,technical aspects and analyst expectations based on the Barra model and related references,including more than one hundred specific subdivision factors.Then this article conducts A-share market distribution law testing on these subdivision factors,including factor difference testing between industries,factor difference testing between companies with different market capitalizations,and correlation testing between factors to determine whether these factors need to be tested in the industry.Neutralize market value.Once again,this article carried out single-factor detection on these candidate subdivision factors.The single-factor detection screened out the excellent factors in each category through regression test,IC value analysis and stratification test.At last.After the selection of excellent factors under the broad category is completed,this article conducts a factor synthesis test on these factors,so far the factor selection and testing work is all over.After obtaining the seven excellent synthetic factors,this article began to build a multi-factor stock selection model.In the process of factor testing,the research found that the performance of different categories of factors are also very different.Therefore,the commonly used equal weight assignment is discarded in factor weighting and the information ratio weighting that can consider both the return rate and the risk is adopted.law.At the same time,in order to further optimize the multi-factor stock selection model,this article also adds a timing strategy to it to reduce the withdrawal rate of the stock selection model and increase the return rate of the model.Finally,according to the results of the backtest,the multi-factor timing stock selection model created in this article achieved a return rate of 700.56% and an annualized return rate of 23.85% during the 10-year backtest period from 2010 to2019.Compared to the CSI 300's excess return rate of 598.77%,the Sharpe ratio of this model is 0.956,the winning rate is 0.615,and the maximum drawdown rate is23.38%.These research results show that the multi-factor timing stock selection model created in this article for factor effectiveness and factor synthesis has strong profitability and reasonable risk control.