Community detection aims to partition a network into internally densely connected groups of nodes. In huge networks, exploring the whole network is computationally very expensive. For applications such as antiterrorism, the spread of disease on networks and viral advertising, researchers are now more interested to find the community surrounding one or a few individuals. In this context, we propose a local community detection algorithm that starts from a seed node and iteratively expands it to reach the community that resembles the most to a real-life community. The expansion process is controlled by a neural network classifier that decides which nodes to add to the community being expanded. This classifier is built using three features quantifying the strength of relationships between nodes and communities. Experiments achieved on Lancichinetti-Fortunato-Radicchi (LFR) synthetic networks and real-world networks from different application domains, proved the high performance of our method as compared to the baselines.
No takes yet. Share an insight, caveat, or question.
Choumane et al. (2020) studied this question.