 This study examined urban highway tunnels as common accident locations and developed methods to predict and analyze traffic conditions following accidents. It found that random forest and BP mural networks performed well in predicting traffic conditions, with the former performing better in terms of robustness and generalization in predicting crash duration. This research provides valuable insights into how to manage traffic operations during accidents in urban highway tunnels. This article was authored by Yang Shen, Chong Jiang Zheng, and Fei Wu.