Research On Quantitative Stock Selection Model Based On Big Data And Machine Learning

With the popularization of computer technology,China's quantitative investment business has developed rapidly.However,compared with the developed country market,it started relatively late.Its related theories and applications have drawn much attention in the investment field in China.Quantitative investment techniques mainly refer to the use of quantitative methods such as mathematical statistics and computerized programmatic transactions to obtain excess returns,and are favored by investors for their systematic,disciplined,and decentralized characteristics.This paper mainly studies the effectiveness of the multi-factor stock selection model applied to the A-share market and further improved the model's performance in stock selection effect by machine learning algorithms.This paper selects 44 factors as candidate factors which are divided into eight categories,including quality factors,talent factors,valuation factors,scale factors,transaction factors,emotion factors,technical index factors and some WorldQuant Alpha101 factors.The time range of the empirical test is from January 1 st,2010 to December 31st,2017.Firstly,this paper selects 10 relatively independent effective factors from 44 candidate factors through validity test and de-redundancy treatment.Secondly,we establish multi-factor model,test its validity and introduce hedging mechanism to further reduce the maximum drawdown of the model.Then,we change the factor lookback period to select the best model for backtesting;Finally,we combine AdaBoost algorithm and multi-factor model to form a new model,and compared it with the traditional multi-factor model in stock selection effect.We found that machine learning AdaBoost algorithm can enhance the stock selection effect of the traditional multi-factor.