Application Research Of LightGBM Algorithm Based On Multi-factor Stock Selection Model

Quantitative investment refers to taking advantage of modern statistical and mathematical methods,and using computer technology for the purpose of securities investment transactions.In the field of quantitative investment,the multi-factor stock selection model is the most widely used,deepest and most researched stock selection model in the domestic quantitative investment field.The core of the multi-factor model lies in the selection of factors.However,due to the frequent changes in the rules of the A-share market in recent years,the traditional method of screening factors has a strong subjectivity and timeliness,making it difficult for the multi-factor stock selection model to predict fitly obtain stable and considerable returns.This paper first sorts out the basic principles of the multi-factor stock selection model and points out the shortcomings of the traditional multi-factor stock selection model in the process of screening of effective factors.Based on above procedure,this paper combines the the multi-factor model with the latest machine learning technology in order to optimize the traditional Multi-factor investment model.In order to further optimize the traditional multi-factor model,this paper selects the most advanced Light GBM algorithm in Boosting algorithm,exchanges the critical step of factor selection in the traditional multi-factor model,and replaces human subjective selection with the algorithm.Therefore,a multi-factor stock selection model based on Light GBM algorithm was proposed.On the basis of the classic multi-factor model,this model in this paper firstly constructs an alternative factor pool;secondly,it uses Light GBM algorithm to select effective factors from the candidate factor pool;and then uses the obtained effective factors after the above operation to predict the stock's rise or fall,and then selects 10 stocks which have the highest probability of rise to build a portfolio for each period.Finally,using the CSI 300 data of China A-share market as sample data to perform an empirical study.The results of the research show that multi-factor stock selection model based on Light GBM algorithm may has better predictability.