Graph reconstruction model for enhanced community detection

Abstract

Community detection is a fundamental task in complex network analysis, focusing on uncovering the underlying organizational structures of networks by analyzing relationships between nodes. While existing methods have shown significant success, they often struggle in networks with overlapping communities or intricate topologies, primarily due to their reliance on local information and limited ability to capture global structures. To overcome these limitations, we introduce the Graph Reconstruction Model for Enhanced Community Detection (GRMECD), a novel approach that integrates higher-order information with network reconstruction. Leveraging a Markov chain-based transfer probability matrix, GRMECD captures the global network structure, enabling effective pruning and reconstruction to enhance the performance of community detection. Experimental evaluations on synthetic and real-world datasets demonstrate that GRMECD consistently outperforms state-of-the-art methods, particularly in networks with complex or overlapping structures.