Constrained Community Detection in Social Networks

TL;DR

This paper introduces a constrained community detection method that integrates external constraints into modularity optimization, demonstrated by identifying hospital communities with standardized cardiac care levels.

Contribution

It proposes a novel penalty-based approach to incorporate constraints into modularity optimization for community detection.

Findings

Successfully identified hospital communities with balanced cardiac care.

Enabled meaningful comparisons of healthcare quality across communities.

Demonstrated applicability to healthcare network analysis.

Abstract

Community detection in networks is the process of identifying unusually well-connected sub-networks and is a central component of many applied network analyses. The paradigm of modularity optimization stipulates a partition of the network's vertices which maximizes the difference between the fraction of edges within groups (communities) and the expected fraction if edges were randomly distributed. The modularity objective function incorporates the network's topology exclusively and has been extensively studied whereas the integration of constraints or external information on community composition has largely remained unexplored. We impose a penalty function on the modularity objective function to regulate the constitution of communities and apply our methodology in identifying health care communities (HCCs) within a network of hospitals such that the number of cardiac defibrillator surgeries performed within each HCC exceeds a minimum threshold. This restriction permits meaningful comparisons in cardiac care among the resulting health care communities by standardizing the distribution of cardiac care across the hospital network.