Detecting communities around seed nodes in complex networks

Abstract

The detection of communities (internally dense subgraphs) is a network analysis task with manifold applications. The special task of selective community detection is concerned with finding high-quality communities locally around seed nodes. Given the lack of conclusive experimental studies, we perform a systematic comparison of different previously published as well as novel methods. In particular we evaluate their performance on large complex networks, such as social networks. Algorithms are compared with respect to accuracy in detecting ground truth communities, community quality measures, size of communities and running time. We implement a generic greedy algorithm which subsumes several previous efforts in the field. Experimental evaluation of multiple objective functions and optimizations shows that the frequently proposed greedy approach is not adequate for large datasets. As a more scalable alternative, we propose selSCAN, our adaptation of a global, density-based community detection algorithm. In a novel combination with algebraic distances on graphs, query times can be strongly reduced through preprocessing. However, selSCAN is very sensitive to the choice of numeric parameters, limiting its practicality. The random-walk-based PageRankNibble emerges from the comparison as the most successful candidate.

Local Community Detection Based on Small Cliques

“…Numerous local community detection algorithms have been proposed [4,[16][17][18][19]. Many of these algorithms can be grouped as greedy community expansion [4], which also provides the basis of other (more complex) algorithms.…”

“…Many of these algorithms can be grouped as greedy community expansion [4], which also provides the basis of other (more complex) algorithms. It can be described best as greedily adding the node to the community that currently has the maximum value of some function that assigns a value to each node.…”

“…In each step, the next node is selected by iterating over all nodes in the shell of the community. We examine the two quality measures proposed for this algorithm by [4]: The M measure [29] and the L measure [17].…”

“…Having access to the whole network requires huge computational resources and is basically impossible, so local community detection algorithms provide a viable alternative here. Various local methods have been proposed that use the structure of the network to detect communities locally [3,4]. The study in [4] however shows that local algorithms still perform significantly worse than global, disjoint community detection algorithms on synthetic benchmark graphs.…”

“…Various local methods have been proposed that use the structure of the network to detect communities locally [3,4]. The study in [4] however shows that local algorithms still perform significantly worse than global, disjoint community detection algorithms on synthetic benchmark graphs. Expanding single nodes (LFM [5], OSLOM [6]), single edges (MOSES [7]) or maximal cliques (GCE [8]) into communities is nevertheless a successful scheme for overlapping community detection.…”