Community Detection (Louvain, Walktrap)

Community detection algorithms such as Louvain and Walktrap partition a network into groups of nodes that are more densely connected to each other than to the rest of the network.

루뱅과 워크트랩 같은 커뮤니티 탐지 알고리즘은 네트워크를 나머지 부분보다 서로 더 촘촘하게 연결된 노드들의 집단으로 분할합니다.

How do I find groups or communities of closely connected nodes in a network?

네트워크에서 서로 긴밀하게 연결된 노드 집단(커뮤니티)은 어떻게 찾나요?

Community detection identifies clusters of nodes with dense internal connections, with Louvain greedily optimizing modularity and Walktrap grouping nodes that short random walks tend to visit together. It fits social, citation, co-occurrence, or similarity networks where the goal is to uncover subgroups from the pattern of ties.

When to use Community Detection (Louvain, Walktrap)

• Social networks where cohesive subgroups are of interest
• Keyword co-occurrence or citation networks to identify themes
• Networks built from a similarity or correlation matrix among variables

When to use something else

• Data are attributes of cases rather than ties between them: use k-means or hierarchical clustering
• Disconnected communities from Louvain are a concern: use the Leiden algorithm

Assumptions and requirements

• Meaningful network with defined nodes and edges (optionally weighted)
• Mainly undirected networks for standard implementations
• Modularity has a resolution limit that can merge small communities
• Louvain results can vary between runs, so stability should be checked

What the analysis reports

• Community membership for each node
• Number and sizes of communities
• Modularity (Q)
• Network plot colored by community

How to report it

The Louvain algorithm identified [k] communities with a modularity of Q = [value].

Related methods

• Leiden Algorithm
• Hierarchical Clustering
• Network Centrality Analysis

Run Community Detection (Louvain, Walktrap) on Quorant