The Adaptability Of Machine Learning Algorithms To Chinese A-shares

In the field of investment,there are many kinds of investment styles,some people rely on basic analysis,suck as Buffett.Some of them rely on technical analysis,Such as John Murphy.In recent years,with the continuous development of computer technology,the European market has already appeared in a quantitative investment trend,and continues to grow,The faction is represented by James Simmons.Quantitative investment win with the advantages of stable,rational and probability,it occupy the main stream of investment.Machine learning is the cutting-edge technology in the field of data mining,many machine learning algorithm with its rigorous logic of training data for feature learning,then learn to predict the characteristics of experience.In the area of investment,successfully predicted the future trend is the best guarantee of successful investment.Many scholars have tried various machine learning algorithm to the field of quantitative investment,some have achieved success,but also failure because of inappropriate investment.The main reason for the failure is that different algorithms have different adaptability to different data.While the stock market data full of myriads changes,different period and different varieties will produce different kinds of data.This paper tries to make a comparison of seven kinds of machine learning algorithms in Chinese A share market,results show that for the price forecast,SVM wins in rising trend.Random forest wins in downward trend,all algorithm perform common in consolidation trend.For the next level of opening forecast,decision tree wins in rising trend,neural network wins in downward trend,logistic regression wins in consolidation trend.For different varieties of the price forecast,Naive Bayesian wins in the industrial and agricultural stocks.Logistic regression wins in service sector,support vector machine wins in high-tech industry stocks.For the next day's prediction of different varieties,the logistic regression algorithm is almost optimal,only in the high-tech industry stock,the SVM algorithm performs better thanthe logistic regression algorithm.