Multi-Factor Quantitative Stock Selection Strategy Based On Random Forest Algorithm

In financial market,how to combine investment and math theory has been a hot topic for scholars and practitioners for a long time.And quantitative investment came into being for that reason.Quantitative investment is the method to realize investment strategy with knowledge of finance,math,statistics and programming.As the result of combining finance and math theory,quantitative investment is strongly systematic and disciplined,and has become very popular in American stock market.Compared with America’s early-start and developed quantitative investment technology,China’s quant yield started late having long way to go.Chinese investors’knowledge and acceptance level of quantitative investment is low thus need more education.Chinese stock market is the world second largest behind American stock market based on market cap.However,when based on proportion of quantitative investment,Chinese stock market is far behind American stock market.Soconsidering development of developed market,our quantitative investment has considerable room for development.The article’s purpose is to improve traditional multi-factor stock selection strategy with random forest algorithm.Among the quantitative investment strategies,multi-factor stock selection is one of the most commonly used ones.But as many years’ development,the easy and useful factors have almost lost their effect.Since 2017,due to the great change of market situation,many factors which used to make huge excess return have been faded,which caused quantitative fund to be challenged.These years,as artificial intelligence get popular,industries have begun to integrate Al technology.AI’s application is sure to deeply influence industries.Meanwhile,people start to consider combining AI with quantitative investment to get more return.It has been proved that machine learning,the base algorithm of AI,has advantages when combined with traditional multi-factor stock selection.But investors and scholars haven’t much realized that.So research about the machine-learning-enforced multi-factor stock selection strategy has great meaning to earn more excess return,develop market size of quantitative investment and increase scale of asset management companies.The article uses WindQuant platform and Python programming language as tool,takes objective to reinforce traditional multi-factor stock selection with machine learn algorithm and meanwhile try to enhance its all aspect performance,improve effectiveness of multi-factor stock selection strategy.Election of effective factors is base of multi-factor election strategy,so that’s where we start.Through factor election,effectiveness testing,we filter out the effective factors we need.Then we extract factor data from WindQuant platform with Python,and fill missing values,neutralize data to make it suitable for our strategy building.Next step is to buildup multi-factor stock selection strategy with random forest algorithm,back test and evaluate the strategy.At the end of article,we make a try to improve the strategy.The article’s main framework mainly contains:factor selection,analysis of factors’ effectiveness,factor filtering,build up of strategy,back test and improvement of strategy,review and evaluation.Through back test,the article proves that in the period between 2014 and 2019,stock list of ZZ500,the article’s quantitative strategy has strong selection ability,and has achieved high return.The strategy not only outperforms ZZ500 index,but also has many advantages compared with traditional multi-factor stock selection.The article’s strategy has strong feasibility and practicality,and has significance for quantitative research and development of stock quant strategy funds.