Weighted Graph Clustering with PaCCo
Arun Kumar Soma · 2025
Graph data analysis benefits from detecting substructures to infer meaning and interpret data correctly. However, many existing graph clustering algorithms are not applicable to weighted graphs, which contain edge weights representing interaction magnitudes. To address this, the PaCCo algorithm maximizes node similarity and cluster interconnectivity while accounting for inter-cluster links. This paper reviews related algorithms and evaluates PaCCo’s performance. Results show that PaCCo offers benefits such as parameter independence, automatic clustering, and reduced runtime, making it suitable for various applications. Future enhancements could include incorporating fuzzy clustering and transitioning to a distributed computation paradigm for scalability.