Resolution controls the hierarchical-level information about community formation. It represents the level in a dendrogram at which to converge for the number of communities to be detected. Think of different resolution points as different hierarchical levels in a tree of nodes interconnected through an edge table.
Two ways to think of resolution are:
|Higher Resolution (> 1.0)
|Provides a visualization of the graph nearer the root of the tree.
|Provides a visualization of the graph nearer the leaves of the tree.
|"Zooms out ," providing fewer, larger communities.
|"Zooms in ," providing more, smaller communities.
In the following figure, the x-axis represents the nodes in a sample graph and y-axis represents the community resolution level of the hierarchical tree structure of the graph. For example, at resolution level 0.75, the graph has 7 communities, and at resolution level 1.0, the graph has 5 communities.
Resolution is useful for:
- Visualizing hierarchical graph structure
- Finding the expected number of communities in a graph
- Finding the community structure in a graph when some nodes are expected to belong to the same community
For visualizing hierarchical graph structure, the default resolution (1.0) maximizes modularity and is therefore expected to provide the best community structure. Possible exceptions are the second and third preceding use cases, or a graph that includes estimation errors (for example, when edge weights are statistically computed based on behavior similarity among nodes).
For the third use case, an alternative to changing the resolution is to put the nodes that are expected to belong to the same community in one group in the vertices table and specify that group in the CommunityAssociation argument.