Estimation Method of Traffic State Parameters Based on Toll Data

To obtain the spatial and temporal traffic operation parameters of a road network more accurately using toll Origin-Destination (OD)data, this paper proposes a method to calculate the traffic operation status by combining machine learning prediction of the service area approach with incremental iterations of toll OD data. The results show that there is a maximum peak value for the minimum probability threshold Pt of entering the service area, and the overall prediction accuracy is highest when Pt is approximately 0.9. In addition, the random forest method shows a significant accuracy advantage in different categories of toll forms and vehicles. The OD data for short-distance trips are used as the initial value to project the travel time of the basic road section, and the road network operation status is calculated incrementally from near to far using the multiples of the distance threshold Pt as the time interval. The OD data calculated in the previous iterations are weighted more heavily than the data in the later iterations to complete the road section status for missing data in the spatio-temporal dimension. The effectiveness of the algorithm was tested in a large-scale provincial real-world road network environment by designing four different sets of Pt and three control groups. The results of the analysis of road traffic flow and travel speed show that the smaller the Pt, the smaller the error. The error is minimized at Pt=5 km, where an average error of 5.46% for traffic flow and 9.84% for speed, using the random forest model. Compared with the nearest neighbor upstream and downstream interpolation method, the average accuracy of the proposed method for calculating traffic flow is 3.98% higher, and the average accuracy of speed is 4.33% higher than that of the average travel time method. The average accuracy of traffic flow is 5.2% higher, and the average accuracy of speed is 5.87% higher. The test results show high accuracy and indicate that the method has good practicality. © 2022, Editorial Department of China Journal of Highway and Transport. All right reserved.