Map Matching of GPS Data with Bayesian Belief Networks

This paper proposes a map matching algorithm using Bayesian belief network for GPS traces to generate the spatial-temporal information of individuals. The algorithm incorporates the road network topology, distance from trace nodes to road segments, the angle between two lines, direction difference, accuracy of measured GPS log point, and position of roads. The GPS data collected in the Eindhoven region, The Netherlands, was used to examine the performance of this algorithm. Results based on a small sample show that the algorithm has a good performance in both processing efficiency and prediction accuracy of correctly identified instances. Even with a small sample, the overall prediction accuracy reaches 87.02%.