Construction And Application Of Multi Factor Stock Selection Model Based On Random Forest

In recent years,the role of markets in allocating resources has been strengthened by the gradual development of our system of securities registration,the liberalization of the restrictions on the ratio of foreign shares and the improvement of the delisting system.The concept of quantitative investment has also gone from the "cognitive" stage to the "familiar" stage.On the basis of mathematical statistics,a quantitative investment model with clear buying and selling point,position period and other conditions and strict implementation can avoid the subjective deviation of investors.Among many trading models,the multifactorial model is the most robust and promptly to quantify investment,and the machine learning transaction model also plays a unique role in exploring the economic laws underlying stock market data.Therefore,on the basis of the traditional multi factor model trading strategy,using machine learning algorithm to adjust and optimize has practical significance,And the initial intention of this paper is to carry out this research.Firstly,this paper describes the theoretical research process of multi factor model in the context of time development,partly combining the application of machine learning in the field of quantitative investments in recent years,so as to make readers have a more intuitive understanding of the theoretical basis of building multi factor stock selection model.Secondly,this paper preliminarily constructs the factor pool from four aspects,and analyzes some of the factors In order to improve its applicability in the A-share market,this paper analyzes the effectiveness of factors from three aspects of profitability,forecasting ability and turnover frequency,as well as the short-term,medium-term and long-term trading cycles,and constructs the initial effective factor pool;thirdly,through the correlation test of effective factors,it selects the factors with high correlation,taking into account the factors In addition to the correlation between factors,there may be more unknown and effective information.This paper constructs seven factor combinations based on the lowest five factors,and tests different factor combinations through random forest algorithm.Finally,five factors are selected to construct the model,including price to book ratio index,relative strength index,smooth moving average of similarities and differences,intermediate willingness index and trading volume standard deviation index Finally,we choose the trading data of A-share market from September 1,2019 to September 1,2020 to simulate the back test of multi factor model,and compare the back test results of different factor combinations.The results show that the multi factor model can overcome the market and obtain the risk adjusted return beyond the CSI 300 benchmark,that is,the multi factor model is still effective and suitable in the A-share market Usability.At the same time,the improved five factor model has higher profitability,anti risk ability and risk adjusted profitability,that is,on the basis of the traditional multi factor model,the multi factor model based on random forest can effectively improve the performance of quantitative trading model.Multi factor model has the characteristics of keeping pace with the times,so investors can not be stubborn in the existing factor combination model,they need to combine with the current market to find the most effective factor.At the same time,investors can give full play to the advantages of artificial intelligence in the financial field on the basis of adjusting and optimizing the model with machine learning algorithm.But investors should not fall into the trap of historical data and be limited to factor mining and portfolio construction.The real effective trading model lies in the deep understanding of the market and the economic law behind the data.