Non-Overlapping Community Detection Algorithms: A Comparative Study on Real-World Networks
Year: 2013
Type: preprint
Abstract: A network has a non-overlapping community structure if the nodes of the network can be partitioned into disjoint sets of nodes such that each set of nodes is densely connected to the nodes inside the set and sparsely connected to the nodes outside it. To uncover the community structure in a network, many methods have been proposed till now. Even though there have been a multitude of comparative studies of these methods, there has not been any comparative study of community detection methodologies that rely only on structural data with methodologies that rely entirely on behavioral data. For this purpose, we execute some behavioral property based community detection methods along with some structural property based methods on two real-world networks: Filmtipset and Twitter. The results are then objectively evaluated using two well-defined quality functions: modularity and like-mindedness. We also present two new behavioral property based community finding algorithms called Modified Louvain Method and Like-Mindedness Maximization. The first algorithm attempts to achieve a community structure with approximately the highest modularity and reasonably good like-mindedness. The second algorithm attempts to maximize the like-mindedness of the identified community structure.
Source: arXiv (Cornell University)
Language: English
Subfield: Statistical and Nonlinear Physics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1736347472