Corporate Misconduct Prediction in the Construction Industry Using XGBoost: An Ensemble Learning Approach

Corporate misconduct in the construction industry may lead to severe economic loss and even fatal injuries to workers and residents. An effective way to detect corporate misconduct timely is needed. This paper provides a tool to predict corporate misconduct by analyzing corporate data from 119 listed construction companies in China. XGBoost is used to construct a prediction model for corporate misconduct, and a support vector machine model is used as a benchmark to evaluate the performance of the built XGBoost model. The accuracy of the built XGBoost model in predicting corporate misconduct is 80.38%, outperforming the support vector machine model. The results may facilitate companies’ stakeholders to predict and identify corporate misconduct timely and accurately, and thus corporate scandals may be nipped in the bud.