Application of supervised learning in the exploring research on employment quality for college graduates

Employment quality is a problem of public administration with comprehensive connotations and complex objective and subjective influencing mechanisms. To explore the factors influencing the employment quality, our study previously uses the linear regression model, which is one of the traditional statistical research methods to test the hypotheses based on the literature study. There are some serious methodological problems in traditional statistical research on employment quality. Therefore, we make a thorough investigation of three major state-of-the-art supervised learning models applied to the same survey data collected from college graduates in China, and find that the tree-based model and kernel-based model perform better than the traditional linear regression model, and especially the Multiple Additive Regression Trees (MART) model provides not only reasonably good performance but also excellent comprehensibility for the problem.