A Stock Prediction And Quantitative Investment System Based On Machine Learning

Quantitative investment,as an intersection of computer science and finance,has drawn much attention in recent years.The stock market is a nature scenario to apply classification and regression,which are two main technologies in machine learning.Therefore,extensive attempts to apply machine learning to investment have been made.However,from the view of the industry,it is not easy to effectively apply machine learning to investment,and there are also many problems to solve.Two key issues in current research and applications are reviewed:(1)the features used in the machine learning algorithm which is used to predict the return of stocks depend too much on technical indicator,which contributes to poor explanatory power and less robust,(2)the machine learning algorithm cannot cope with the sudden changes of the market style.This paper designs and implements an investment system,and puts forward a solution to the above two problems.In order to solve the problems in feature selection,a framework for feature testing is designed.Then,a database of features is built on the basis of the domain knowledge.Some features are selected from the database,with which a classifier with Adaboost algorithm to predict the future return of stocks is trained,and an investment system that can invest by itself is constructed.In this investment system,features can be monitored,and the explanatory power is enhanced.In order to deal with the sudden changes of the market style,an advanced method that can choose time points in the same market style as training set with the Hidden Markov Model is put forward.The results of the back-test and the production environment test show that the modified system is able to capture the changes of the market style.During the test period,the modified system is found to have a better performance than the unmodified one.At present,the system implemented in this paper is used as the core backend module in an official launched intelligent investment advisory APP called Zhiyu Liangtou,which provides helpful suggestions on investment for retail investors.In the future,this system is expected to provide support for more client-side products.