Enhancing construction safety management through multivariable grey model analysis and variable selection optimization

In this study, a multivariable grey model (GM(1, N)) is employed to explore how different combinations of variables impact the accuracy of construction accident prediction, using a full permutation algorithm. The aim is to optimize variable selection and improve prediction accuracy. By conducting an exhaustive analysis of 511 potential combinations involving nine variables, it was observed that by integrating crucial external variables such as macroeconomic indicators and industry scale, the multivariable model achieved a prediction accuracy error rate of less than 0.5%, thereby significantly enhancing its information capture and forecasting precision. The analysis suggests that optimal predictive performance is achieved when the number of control variables is approximately four. Additionally, further research shows that increasing the dataset size significantly enhances the model’s predictive capability. This study highlights the scientific rigor and precision of decisionmaking in preventing construction accidents and provides empirical evidence for construction safety management. The research in this paper not only enriches the connotation of the grey system prediction model theoretically, but also provides a data-driven decision support tool for urban construction and safety accident prevention in practice.