Improving the DF-Louvain Algorithm through Random Walk-Based Refinement

Dynamic Frontier Louvain updates communities efficiently by restricting local moves to nodes affected by graph changes. This work adds a random-walk refinement for cases in which deleted internal edges weaken a community enough that it should be split.

The refinement proposes a binary partition within an affected community and accepts it only when the modularity change is positive. This preserves the local-update motivation of DF-Louvain while adding a mechanism for recovering community splits.

Project overview · Code · DOI

Recommended citation: D. H. Do, D. Nguyen, and T. H. D. Phan, “Improving the DF-Louvain algorithm through random walk-based refinement,” in Proc. 2025 RIVF International Conference on Computing and Communication Technologies (RIVF), Ho Chi Minh City, Vietnam, 2025, pp. 932–937, doi: 10.1109/RIVF68649.2025.11365046.
Download Paper | Code