Research On Quantitative Stock Selection And Effect Evaluation Based On Machine Learning

In recent years,with the rapid development of the global economy and the stock market,stock investment has become one of the most commonly used financial management methods.The emergence of quantitative investment provides investors with new investment tools.With its advantages such as discipline,accuracy,timeliness and systemicity,it has been favored by institutions and individual investors.In recent years,with the rapid development of artificial intelligence technology,machine learning has been used by more and more scholars in quantitative investment research.In the empirical study,this paper takes the components of CSI 300 index from January 1^st,2013 to December 31^st,2019 as samples.The data from January 1^st,2013 to December 31^st,2017 was used for machine learning model training and hyperparameter adjustment,and the data from January 1^st,2018 to December 31^st,2019 was used for quantitative stock selection strategy research.The main research contents are as follows:?1?Comparison of the prediction capabilities of the five machine learning models on the future direction of the stock price.?2?Comparative analysis of the backtesting effect of the quantitative strategy based on the different machine learning model;?3?Based on the average distribution.?3?Comparative analysis of quantitative strategies based on equal distribution and equal weight distribution.Firstly,the principal components analysis method is used to reduce the dimensionality of the selected 23 features.The five machine learning algorithms of support vector machine,logistic regression,K-nearest neighbors,random forest,and XGBoost are used to predict the direction of the stock price's future 15 days.The empirical results show that the random forest and XGBoost have the best prediction performance,with accuracy rates of 69.76%and 78.51%,respectively.The accuracy of support vector machines and K-nearest neighbors is about 60%.Logistic regression has the worst prediction performance,with an accuracy rate of only 55.62%.Secondly,the five machine learning models are applied to the research of quantitative stock selection strategy.Through backtesting,it is found that the quantitative strategy based on logistic regression,XGBoost and random forest model has a good performance.Limiting positions by buying stocks with a higher probability of rising has no obvious advantage over unrestricted positions.Random forest and XGBoost perform better in the average allocation of funds.Support vector machines and K nearest neighbors have poor returns in the two types of capital allocation.Logistic regression has no obvious difference between the two types of capital allocation.