Dynamic group-finding in public bus networks

Group structure detection presents an insight of the potential organization and functional properties, and benefit the data packet propagation in the various networks. This paper finds that the average number of neighbors of bus changes in a regular way across the running time. Thus we present a dynamic group-finding algorithm in public bus networks. The proposed algorithm includes two phases. To begin with, time is divided into some time interval based on the average number of neighbors of bus, and the time interval is extracted from the bus mobility trace. A group detection algorithm based on spectral clustering is then proposed for each time interval. The simulation results over a realistic bus trace data show that our dynamic group-finding algorithm well adapts to public bus networks.