Date of Completion


Document Type

Honors College Thesis



Thesis Type

Honors College

First Advisor

James Bagrow


networks, link communities, overlapping communities


One way to analyze the structure of a network is to identify its communities, groups of related nodes that are more likely to connect to one another than to nodes outside the community. Commonly used algorithms for obtaining a network’s communities rely on clustering of the network’s nodes into a community structure that maximizes an appropriate objective function. However, defining communities as a partition of a network’s nodes, and thus stipulating that each node belongs to exactly one community, precludes the detection of overlapping communities that may exist in the network. Here we show that by defining communities as partition of a network’s links, and thus allowing individual nodes to appear in multiple communities, we can quantify the extent to which each pair of communities in a network overlaps. We define two measures of community overlap and apply them to the community structure of five networks from different disciplines. In every case, we note that there are many pairs of communities that share a significant number of nodes. This highlights a major advantage of using link partitioning, as opposed to node partitioning, when seeking to understand the community structure of a network. We also observe significant differences between overlap statistics in real-world networks as compared with randomly-generated null models. By virtue of their contexts, we expect many naturally-occurring networks to have very densely overlapping communities. Therefore, it is necessary to develop an understanding of how to use community overlap calculations to draw conclusions about the underlying structure of a network.

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.