Calibration of traffic dynamics models with data mining

Speed-density relationships are one of models used by a mesoscopic traffic simulator to represent traffic dynamics. While the classical speed-density relationships provide a useful insight into the traffic dynamics problem and have theoretical value to traffic flow, for such applications they are limited This paper focuses on calibrating parameters for the speed-density relationships by using data mining methods such as locally weighted regression, k -means, k -nearest neighborhood classification and agglomerative hierarchical clustering. Meanwhile, in order to improve the precision of the parametric calibration, we also utilize densities and flows as variables to calibrate parameters. The proposed approach is tested with sensor data from the 3^rd ring road in Beijing. The test results show that the proposed algorithm has great performance on the parametric calibration of the speed-density relationships.