A new approach to finding overlapping community structure in signed networks based on Neutrosophic theory

Abstract

Abstract One of the longstanding challenges in network science is the identification of overlapping community structures. Real-world networks often exhibit a complex interplay of positive and negative relationships, making the recognition of overlapping communities a crucial area of research. Current community detection methods in signed networks primarily focus on discovering disjoint communities, where each node belongs exclusively to a single community. However, these algorithms often fail to detect overlapping communities, where nodes can belong to multiple communities simultaneously. To address this limitation, we propose a novel approach called Neutrosophic c-means Overlapping Community Detection (NOCD) based on neutrosophic set (NS) theory. By incorporating NS theory, our approach effectively handles the uncertainty associated with ambiguous community boundaries and appropriately handles nodes on the community boundaries and isolated nodes. The NOCD method comprises two phases: firstly, a signed graph convolutional neural network is employed to learn the structural features of the signed network in a lower-dimensional representation; secondly, overlapping communities are detected using the neutrosophic c-means algorithm applied to the embedded network. To evaluate the effectiveness of our proposed NOCD method, we conducted comprehensive experiments on both real and artificial networks. The experimental results demonstrate the effectiveness and robustness of NOCD in identifying overlapping communities, outperforming existing methods. [Received on 3 August 2023; editorial decision on 13 December 2023; accepted on 19 December 2023]