Parameters Calibration of Traffic Simulation Model Based on Data Mining

Aiming at the limitation of the classical speed-density model on describing the dynamic change characteristics of the traffic flow, this paper put more road detected information in the process of parameters calibration of traffic simulation model. After preprocessing the detector data, the data mining methods are used to calibrate the vehicle speed. It also proposes a novel locally weighted regression based on agglomerative hierarchical cluster. It first clusters the training samples and uses the agglomerative hierarchical clustering algorithm to establish a clustering tree for each constraint-clustering. Then it applies the k-nearest neighbor method to cluster new stage samples into the best fit clustering. Finally, the vehicle speed is estimated. The vehicle density, densities and flows are taken as the variables. The test with a huge of field data shows that the proposed algorithm performs well on parameters estimation precision and efficiency. It is appropriate for DTA based simulation.