Using Bi-partite graphs to cluster complex networks

Kaine Black, Monica Wachowicz, Alec Parise

Abstract: The Internet of Things (IoT) has been well established as the next major innovation in the internet and connected devices. The IoT will consist of static sensors, sensors that remain in a fixed location, as well as mobile sensors, sensors that are in motion during their operation. These IoT devices will have to communicate with one another in order to achieve a common goal, and make smart decisions. How these complex networks of devices determine which other devices they will communicate with is the problem addressed in this paper. The methodology used here consists of three major components; the construction of mobility neighborhoods, the utilization of bipartite graphs to represent the network, and the clustering of the bi-partite graph using the Louvain community detection algorithm to partition the network into communities of high modularity. This methodology is implemented using a dataset based on vehicle trajectories on a 600 meter strip of Highway 101 in the United States. The preliminary results show that the methodology can be used to find clusters of vehicles with a high modularity along the strip of highway. These results are preliminary and case specific to the vehicles on the highway, however the general methodology could be applied to any network of IoT devices.