Governance of executive personal characteristics and corporate performance based on empirical evidence based on machine learning

In the current corporate governance literature, most studies on executive characteristics only focus on the relationship between a single executive characteristic and company performance, and lack a comprehensive analysis of executive characteristics; on the other hand, they mainly focus on causal inference. This article uses the Boosting regression tree algorithm in machine learning to thoroughly investigate the relationship between the company's multidimensional performance characteristics and performance, while avoiding the weaknesses of traditional linear models and is more suitable for analyzing nonlinear and interactive relationships between variables. This article uses listed companies from 2015 to 2020 as a sample to empirically evaluate the ability of executives to predict company performance, further dig out the personal characteristics of executives with strong ability to predict company performance, and describe their prediction mechanism. Experiments have shown that the cross-term coefficient of the proportion of senior management's shareholding and age is - 0.028, showing a significant correlation (P = 0.005 < 0.05), and the cross-term coefficient of the proportion of senior management holdings and education is - 0.003, showing a significant correlation (P = 0.005 < 0.05), which shows that the management analysis of executives' personal characteristics and corporate performance based on machine learning is more accurate than other traditional technical analyzes, which also opens the future corporate governance performance a new direction of research, and the use machine learning and deep learning algorithms for governance decisions can also reduce the involvement of human factors in future corporate performance governance.