Abstract:

In order to explore the mechanism and key influencing factors of the safety competency of construction workers, this study established a multi-indicator dataset based on organizational network and individual behavior, referring to social capital and behavioral motivation theory. A safety competency indicator system was constructed based on previous scales, and data were obtained from 457 construction workers in China through a questionnaire survey. The multi-indicator comprehensive evaluation method based on the entropy weight and coefficient of variation method was used to calculate and screen the indicators. Then, five machine learning methods, such as GBDT, were used to establish regression models, respectively. The grid search method was used for hyperparameter optimization to improve the prediction ability of the model. The results show that the ensemble learning method is superior to the traditional machine learning method, and the GBDT algorithm has the best fitting performance. The optimized dataset can improve the prediction accuracy of the model. The safety participation and the structural and cognitive dimension of social capital significantly affect the construction workers' safety competency. In addition, their working experience and education level also have a correlation with their safety competency.

Key words: machine learning, construction workers, safety competency, prediction model, gradient boosting decision tree (GBDT)