Research On Quantitative Investment Stock Selection Based On Machine Learning

Quantitative investment can be traced back to the 1950 s and has a history of more than 60 years.With the rapid development of computer technology and big data technology,quantitative investment has once again become a hot spot in the investment community.Quantitative investment can be divided into six types:quantitative stock selection,quantitative timing,stock index futures,commodity futures,statistical arbitrage and asset allocation.The main research work of this paper will focus on two parts: quantitative timing and quantitative stock selection.Quantitative timing can predict the future trend of the stock market and provide information for investors to make investment decisions,while quantitative stock selection can screen out the stocks with the greatest rising probability for investors and obtain stable returns.The combination of the two can provide investors with a low-risk and stable return investment decision.This paper focuses on China's A-share market from 2010 to 2018,constructs a quantitative timing model based on SVM algorithm and a quantitative multi factor stock selection model based on xgboost,and integrates the two models to form a quantitative investment strategy,hoping to obtain stable excess investment income through this strategy.In addition,factor IV is introduced for the first time in this paper.IV value is mainly used to evaluate the prediction ability of variables in the risk control model,and then carry out feature screening.In the field of quantitative investment,no scholars have used IV value to carry out feature screening.In this paper,IV value is introduced into quantitative investment for the first time to screen effective factors,hoping to improve the interpretation ability of effective factors,making the model more intuitive and easy to understand.The empirical results show that by calculating the IV value of the feature,it has more advantages in feature selection,which can effectively improve the performance of the model and improve the prediction accuracy.Therefore,it is feasible and effective to use IV value in feature selection in the modeling process.However,the quantitative investment strategy based on quantitative timing and multi factor stock selection obtained 33.10% excess return and 34.20% excess annualized return in the back test from January 2,2018 to December 28,2018,far higher than the benchmark(CSI 300 index)return in the same period.This shows that the integrated quantitative investment strategy based on SVM timing and xgboost multi factor stock selection is feasible and effective.Compared with the traditional quantitative multi factor stock selection model,when the market is poor,adding the quantitative multi factor stock selection model will have more advantages.Therefore,the quantitative investment comprehensive strategy constructed in this paper can provide some reference for investors to obtain excess return.