Abstract
Community detection problem in networks has received a great deal of attention during the past decade. Most of community detection algorithms took into account only positive links, but they are not suitable for signed networks. In our work, we propose an algorithm based on random walks for community detection in signed networks. Firstly, the local maximum degree node which has a larger degree compared with its neighbors is identified, and the initial communities are detected based on local maximum degree nodes. Then, we calculate a probability for the node to be attracted into a community by positive links based on random walks, as well as a probability for the node to be away from the community on the basis of negative links. If the former probability is larger than the latter, then it is added into a community; otherwise, the node could not be added into any current communities, and a new initial community may be identified. Finally, we use the community optimization method to merge similar communities. The proposed algorithm makes full use of both positive and negative links to enhance its performance. Experimental results on both synthetic and real-world signed networks demonstrate the effectiveness of the proposed algorithm.
Highlights
Many complex systems in the real world can be modeled as networks[1]
The challenge of the community detection problem in signed networks is that the community structure is ambiguous since that there are some negative links within communities and some positive links between communities
Several algorithms have been extended from the community detection algorithm in unsigned networks to solve community detection problem in signed networks[20]
Summary
Many complex systems in the real world can be modeled as networks[1]. The networks that include only positive links are called unsigned networks, and the networks with both positive and negative links are called signed networks. Community detection problem has attracted increasing attention since it was first proposed by Girvan and Newman[9] Most of these community detection methods can only handle the networks without negative links, i.e. unsigned networks[9,10,11,12,13,14,15,16,17,18,19]. It uses the GN algorithm to detect communities based on the positive links, and combine the negative links to get the final hierarchical clustering results. We propose a random walk-based algorithm named SRWA for community detection in signed networks based on positive and negative links. If a node could not be added into current communities, a new initial community may be developed Experimental results on both synthetic and real-world signed networks show the feasibility and effectiveness of the proposed algorithm
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
