Cluster-based network proximities for arbitrary nodal subsets

Area of Science:

• Network science
• Data analysis
• Computational social science

Background:

• Network analysis frequently involves identifying clusters or communities.
• Existing methods for community detection and clustering vary in complexity and applicability.
• Understanding network structures is crucial in diverse scientific domains.

Purpose of the Study:

• To introduce a unified measure of cluster-based proximity between network nodes.
• To provide a flexible approach applicable to various network types and subsets of interest.
• To enhance the dialogue on defining network clusters and communities.

Main Methods:

• Utilizing random walks and geodesic distance to quantify node proximity.
• Developing a unified measure applicable to specific subsets of nodes within a network.
• Comparing the proposed method with existing approaches using metrics like Normalized Mutual Information (NMI) and Adjusted Rand Index (ARI).

Main Results:

• The proposed unified measure effectively quantifies cluster-based proximity.
• The method demonstrates applicability to diverse datasets, including multipartite networks.
• Community detection is presented as a limiting case of the broader clustering approach.
• Performance comparisons show favorable results against established methods.

Conclusions:

• The novel proximity measure offers a simple yet informative tool for network analysis.
• This approach can aid in identifying clusters with shared attributes, relevant for targeted interventions (e.g., in health-related social networks).
• The unified framework expands the conceptual understanding of network clusters and communities.