| With the continuous development and expansion of China's social economy,the “new three boards” have already occupied the major part of the current capital market.As of the end of 2017,the market has already included tens of thousands of listed companies,compared with the Shanghai and Shenzhen Stock Exchanges.The total number of companies is even larger.The gradual expansion of the "new three boards" application has a very important impact on the current financing process of some small and medium-sized enterprises in China.Although the application of "new three boards" has become more and more extensive,China still has a lot of deficiencies in the current market.There is not much research on the new three boards.In addition,there is less research on the use of machine learning methods in management to build models of executive team characteristics,investment status,and performance.Therefore,using the data of the new-three-board listed companies to study how to build the model of the characteristics of the senior management team,investment status and performance has a certain practical and theoretical significance.This paper proposes the use of machine learning methods to build a performance prediction model for a new-three-board company.This study selected 10,135 newthree-board companies as research samples,extracted feature variables,and used five methods including multiple-variable linear regression,polynomial regression,support vector regression(two types)and XGBoost to build a prediction model.All these models are evaluated to select the best predictive model.The content of this article mainly includes the following four parts:(1)Construct a corporate performance forecasting model based on the characteristics of senior management in new-three-board companies.The independent variables selected for this experiment are the characteristics of the background of senior executives(age,gender,education),and the fierce characteristics(shareholding ratio)as the characteristics of the senior management team.The dependent variable is the average profit of the company for the past four years.(2)Construct a corporate performance forecasting model based on the investment status of the new-three-board companies.The independent variables selected for this experiment are the number of subsidiaries invested by companies,the number of subsidiaries distributed,the average percentage of shares held by subsidiaries,and the proportion of investment relations.The dependent variable is the average profit of the company for the past four years.(3)The construction of the future investment forecasting model of new-threeboard companies.This experiment focuses on the number of companies that will be invested by the new-three-board companies in the next year and the company's distribution.For this goal,the selected independent variable is the number of companies that the company invests in each year,and the dependent variable is the number of companies that will invest in the next year.(4)Select ten companies listed on the New Three Board,Main Board,and Growth Enterprise Board,and use the best performance models from the first three experiments to analyze and compare the effectiveness of this model in the listed companies except for the New Three Board.The experimental results show that the model constructed by XGBoost method has the best performance.It can not only provide certain scientific support for the company in building the investment of the senior management team and subsidiaries,but also can provide data assistance for the funders to formulate investment strategies. |
