OCDAD: An Overlapping Community Detecting Algorithm using Attention Degree in Directed Ex-EgoNet

Detecting community structure is of great theoretic significance and application value to extracting latent information in complex networks. Various algorithms of community detecting in directed networks have been proposed from a macroscopic scale. However, only a few of them are designed for handling directed networks microscopically. In this paper, we first extended micro network model EgoNet to ex-EgoNet. Then the attention degree was calculated by considering impact of direction edges, common neighbors (Alters) and indirect neighbors (ex-Alters). Next, a novel algorithm from a micro view was proposed, named as OCDAD (overlapping community detecting using attention degree in directed networks). Last, the OCDAD was compared with Infomap and OSLOM by evaluating them on several real world networks. The experimental results demonstrated that OCDAD outperformed Infomap and OSLOM.