Comparative Analysis of Community Detection Algorithms in Large-Scale Financial Transaction Networks

Understanding complex financial systems is crucial for effective regulation and risk management. Network analysis offers a powerful tool to uncover hidden patterns in these systems, however, its application to large-scale transaction data remains underexplored. This paper addresses this gap by applying network analysis techniques to a novel dataset of financial transactions, comprising ۴,۲۸۷ vertices and ۲۷,۸۹۰ edges. A comprehensive evaluation of four prominent community detection algorithms Louvain, Fast Greedy, Infomap, and Label Propagation was conducted. These algorithms were assessed based on the network's topological features, with a focus on their performance in the context of financial data. This analysis revealed that Louvain and its variants, specifically Louvain+ and parallel Louvain, outperformed other methods in terms of modularity, conductance, and computational efficiency. These algorithms demonstrated superior ability in identifying fine-grained community structures, particularly excelling in the initial step of the analysis.

کلیدواژه ها:

complex network analysis ، community detection ، transaction networks ، modularity