Research Of Multi-factor Modelfor Stocks Selection Based On Three-layer Filtering Method

Multi-factor model is one of the current international mainstream quantitative investment model, is also a hot issue in the field of quantitative investment in China at present. Multi-factor model try to explain the stocks’ price and select the best stock stock portfolio.The research of multi-factor model has a certain guiding significance to securities firmsand investment funds operation.In current,most domestic popular quantitative investment modelsare established on the basis of Multi-factor model framework,whichis one of most important issuein quantitative investment fields.This article is on the basis of predecessors’ research results,andstill belongs to the empirical research on the method.At first respectively defined long-term factor, medium factor and short-term factorsaccording to the influence the length of the stock cycle. Next,look for the candidate factors from the aspects of market experience and economic logic. And then screen out each kind’s effectivefactor through the single factor test process.Finally introduced the corresponding identification model of artificial intelligencebased on the characteristics of each kind of effective factors to bulid a stock portfoliowhich has a positive return to realize the alpha hedge.Using data to verify the model,the model prediction accuracy can reach above 70%.The Alpha investment strategy’s annual earnings is 58.5% based on the the model,at the same time,the investment strategy’s maximum retracement is 6.37%.Compared to the traditional multiple factor,not only the risk of the proposed model has been effectively controlled, and the yield is improved significantly.In this paper, the research is significative in the followingtwo aspects. On the one hand, through theproposed model to build performance stock portfolios in the stock market and short the csi 300 stock index futures,which makes it possible to produce stable earningsnot only in the cow but also in the bear market.This proposed model can provide a reference or suggestion for investors.On the other hand,This proposed modelcanexpand the researchers’ thought to do further study.