Traffic state information is critical for drivers’ route choice and traffic management, and there existed numerous studies targeted on traffic state evaluation for expressways. However, few studies have conducted traffic state prediction for expressways, and the congestion mechanism has not been explored comprehensively, i.e., the key congestion contributing factors and their influence sizes. Thus, this study aimed to conduct traffic state prediction and comprehensively explore the congestion mechanism. Initially, the section volume and space mean speed were collected by Closed Circuit Television (CCTV) system in 1-min interval. Based on the Traffic State Index (TSI), the traffic states were divided into four categories, i.e., smooth, relatively smooth, relatively congested, and congested. To capture more congestion contributing factors, weather conditions and geometry conditions were also collected. Before modeling, a variables selection method including random forest and correlation analysis was proposed to solve the huge dimension scale in modeling. Then, an ordered logit model was built, which could consider the traffic states’ sequence properties, to quantitatively analyze the impact of congestion contributing factors and predict the appearance probabilities of traffic states. Meanwhile, the Support Vector Machine (SVM) model was also applied in traffic state prediction to provide a comparison. The results revealed that the appearance probabilities of traffic states are significantly influenced by traffic conditions and their spatial–temporal variations. The foggy weather and the change of lanes can also significantly affect traffic state. Besides, the ordered logit model’s prediction accuracies for training set and testing set were 81.79% and 81.35% respectively. However, the SVM’s prediction accuracies were notable lower than the ordered logit model, which indicated the importance of considering traffic states’ sequence properties. This study can be applied in active traffic management systems to adjust strategies in advance and improve traffic efficiency.

暂无