Research on Traffic State Classification Based on Bound—Constraint Support Vector Machines

According to the characteristics of urban road traffic flow, aim at the state of traffic flow with frequent traffic congestion, based on the theory of three-phase traffic flow, an automatic traffic congestion identification algorithm with single section is proposed. According to the three-phase traffic flow theory, combined with the simulation data of Vissim, the traffic operation state is divided into four states. Traffic volume, speed and road space occupancy were selected as parameters and the traffic state classification model is trained by using bound-constraint support vector machines. By comparing the accuracy of the classification model with different regular coefficients and kernel functions, the most suitable regular coefficients and kernel functions are obtained. The results show that the classification model has a classification accuracy of more than 90%, and the classification characteristics of clusters in different traffic states are obvious. The algorithm can quickly and effectively distinguish the urban traffic state and provide a basis for urban traffic congestion warning and traffic guidance strategies.