A QLearning Approach for an efficient Estimation Of Initial Number of Communities in social networks

Abstract

In modern society, social networks have become the main means of communication, enabling links to be created and facilitating the exchange of information between individuals. Consequently, the study of relationships and their impact on individuals, through social network analysis (SNA), has attracted a great deal of attention from researchers. In particular, communities, which play an essential role in the flow of information within a network, have been the subject of extensive study in the field of social network analysis. The difficulty with any community detection approach lies in the initialization stage, as the number of communities is unknown. This paper presents QESN, a new initialization approach for community detection based on QLearning, a reinforcement learning algorithm, which estimates the initial number of communities, thus providing an initial population of good quality, which is therefore improved using the Brain Storm Optimization metaheuristic. Through empirical studies conducted on seven social networks, we demonstrate the effectiveness of our approach, which outperforms state-of-the-art algorithms.