Identification and Prediction Model for Traffic Blockage State of Highways to Xizang

To identify the traffic blockage state on highways to Xizang in extreme environments, comprehensive evaluation indicators for traffic blockage state based on the entropy-weighted TOPSIS method were proposed by using the traffic blockage event parameters observed on four major highways to Xizang: Sichuan‒Xizang Highway, Yunnan‒Xizang Highway, Qinghai‒Xizang Highway, and Xinjiang‒Xizang Highway. The K-Medoids clustering algorithm was used to categorize the traffic blockage state into different levels. By considering factors such as hazard types, road types, traffic volume, and vehicle type ratios, a classification model for the traffic blockage state on highways to Xizang was constructed based on machine learning algorithms. The results show that the Qinghai‒Xizang Highway has the highest average blockage duration, length, and severity. The Sichuan‒Xizang Highway, while having a slightly lower duration compared to the Qinghai‒Xizang Highway, has a shorter average blockage length, resulting in a lower overall severity of blockage events. Compared with the Yunnan‒Xizang Highway, the Xinjiang‒Xizang Highway has a higher average blockage duration, yet both have shorter traffic blockage length, leading to lower blockage severity. The identification model based on entropy-weighted TOPSIS and K-Medoids clustering can effectively distinguish between different levels of the traffic blockage state on highways to Xizang. The LightGBM algorithm achieves the highest accuracy on the test set, with an accuracy rate of 96.5%. The results indicate that due to differences in geological terrain, climate, traffic volume, and primary functions of each route, there are differences in the traffic blockage characteristics. The model proposed effectively classifies and predicts the traffic blockage state on highways to Xizang with promising accuracy.