Predicting hazard degree levels of metro operation accidents based on ordered constraint Apriori-RF method

Abstract

To explore the non-linear relationship between risk sources and the hazard degree levels of accidents, and to precisely predict the hazard impact of metro operation accidents, we propose the ordered constraint Apriori-RF method for forecasting metro operation accident hazard degree levels. Firstly, the hazard degree of metro operation accidents is quantified from three dimensions: casualties, train delays, and facility damages. K-means clustering is then applied to categorize hazard degree levels. Secondly, the ordered constraint Apriori algorithm is employed to mine valid association rules between metro operation risk sources and accident hazard degree levels. These valid association rules are subsequently employed in the random forest algorithm for training, establishing a reliable and accurate prediction model. Finally, the method is validated using metro accident data from a city in China. The research results indicate that the ordered constraint Apriori-RF method enhances the effectiveness of association rule mining by 74.9% and exhibits higher computational efficiency. The predicted values of the ordered constraint Apriori-RF method have small errors. Compared to traditional Random Forest algorithms, the root mean square error is reduced by 14%, and the weighted root mean square error is reduced by 36%, demonstrating the higher accuracy of the ordered constraint Apriori-RF method and its clear advantages. The research findings provide a precise and effective method for quantitatively predicting the hazard degree levels of metro operation accidents, holding significant theoretical and practical value in ensuring metro operation safety and implementing accident mitigation and prevention measures.