This study, based on 2011–2020 China’s listed companies on GEM as research samples, introduces the BPNN (BP neural network) and GBDT (Gradient Boosting Decision Tree) model into the research of the relationship between internal governance and earnings management, which will be comparatively analyzed with the empirical results of the traditional multiple linear regression model, so as to study its validity and predictive power in the earnings’ management research field. The results show the following. (1) The matching effect of the multiple linear regression model is poor in the analysis of GEM, with a high rate of experimental data distortion. However, the prediction ability of BPNN and gradient lifting tree model is much better than that of the multiple linear regression model. (2) The gradient lifting tree model is comparatively more suitable for the study of accrual earnings’ management, while BP neural network is more suitable for the study of real earnings’ management. Through the above research, new ideas will be provided for the application research of machine learning in the future.

No abstract

No abstract