Multi-factor Quantitative Stock Selection Plan Planning Based On Random Forest Algorithm

In the financial market,how to obtain a higher rate of return has always been a matter of concern for investors and speculators.Quantitative investment theory has become a mainstream foreign investment strategy because of its high degree of integration with mathematics and its high degree of discipline and system.In recent years,with the advent of artificial intelligence,machine learning has gradually entered the field of scholars.Among them,decision trees and neural networks have been used by some scholars to solve economic problems,and more machine learning algorithms and quantitative investment Combination is a new trend in the future.In the history of quantitative investment in the United States for more than 30 years,the volume of quantitative investment in the current US secondary market accounted for nearly 80%.Fund managers built quantitative models by mining market information,and then selected stock portfolios to earn pots.full.Among them,the world's famous Renaissance technology company's "Grand Medal Hedge Fund" has an average annual return rate of 34% and is stable.Compared with the mature and complete quantitative investment market in the United States,China's quantitative investment is relatively late,but its development speed is rapid,and there are many development spaces that can be tapped and upgraded.According to the report at the end of December 2017,China's A-share market has ranked second in the world,ranking first in terms of daily turnover,that is,liquidity.But at the same time,in the current situation of the world's hedge funds totaling 3 trillion,of which more than 30% use quantitative investment methods,China's quantitative investment scale is less than 5%.However,the effectiveness of today's domestic A-share market is not strong.Quantitative investment can use a large amount of data statistics and mining to capture market micro-transaction opportunities,enjoy the excess returns of the market brought by the quantitative stock picking strategy and the overall increase of the market itself.The benefits brought.Therefore,it is of great practical significance to select appropriate methods and models to establish quantitative investment strategies to open up the market,and to increase the scale of asset management and obtain higher returns for brokers and related institutions.In the strategy of using the multi-factor model for quantitative stock selection,factor selection and factor classification selection are two key points.In both directions,this article has been optimized accordingly.In terms of factor selection,the data of the factors selected in this paper contains 70 factors related to finance,valuation,momentum,etc.,and there is a certain expansion in the number and types of factors.In terms of factor classification,this paper uses a random forest algorithm to classify factors.After training with the random forest algorithm,it can output the importance of the feature and detect the correlation between the features.Since the model is created using unbiased estimation,it gives the model a better generalization ability,simple implementation,and fast training.In addition,this paper also compares the random forest algorithm with the SVM algorithm,which further highlights the advantages of the random forest algorithm in fault tolerance and avoiding over-fitting.According to the above design ideas,this paper first collects 70 factors that may affect the stock volatility in terms of fundamentals,policy,market and so on.Selecting some A-shares listed before December 31,2008 and the ShanghaiShenzhen 300 Index constituents to construct a multi-factor stock selection model based on the random forest algorithm,and back testing after optimization and modification,and obtained extremely high returns.Rate,with good results.In this paper,the quantitative stock selection strategy designed by combining the new factor selection and the random forest algorithm has obtained higher yields than the market and related indices and strategies.It has certain feasibility and practical significance,and at the same time,the existing stock selection.Strategies and fund companies provide new ideas for the design and development of future stock picking strategies.