Community detection is a fundamental optimization challenge that investigates the identification of communities within graph-structured networks. Although numerous algorithms have been proposed for this problem, many of them are not scalable to large-scale networks and suffer from high computational costs. In this paper, we propose a multi-agent reinforcement learning (MARL) algorithm for community detection in complex networks, which demonstrates superior performance compared to several well-known baseline methods. The proposed approach is evaluated using multiple performance metrics, including majority accuracy and Nautical mile (NMI), and the results indicate strong and competitive performance. Interactive network-based methods are widely applied across various scientific domains, including social sciences and health informatics, where they facilitate the analysis of behaviors and structural patterns. Furthermore, community detection in dynamic networks can benefit from reinforcement learning and local optimization techniques to effectively manage evolving entities. This type of analysis provides a more efficient framework for examining continuously growing and evolving networks.

Keywords: Complex networks Community detection Multi agent systems Reinforcement learning Majority accuracy