The machinery of the weight-based fusion model for community detection in node-attributed social networks

The weight-based fusion model (WBFM) is among the simplest and most efficient ones for modularity-driven community detection (CD) in node-attributed social networks (ASNs) that contain both links between social actors (“structure”) and the actors’ feature vectors (“attributes”). Roughly speaking, the WBFM first converts the attributes into an attributive network so that one obtains the two networks—structural and attributive—instead of the ASN. Then, the two networks are fused into a composite one that is believed to contain the information about both the structure and the attributes and that can be already fed to traditional modularity-driven graph CD approaches. While the WBFM is widely used, it has been understudied analytically and had only a heuristic ground. In this paper, we disclose the mathematical machinery of the WBFM by revealing the objective function of the corresponding optimization CD process and establishing its connection with the traditional ASN CD quality measures. We also propose a pioneering non-manual parameter tuning scheme that provides the desired impact of the structure and the attributes on the CD results within the WBFM. Based on our theoretical results, we further present a well-tunable Leiden-based ASN CD algorithm that declares itself fast and accurate in our multiple experiments with synthetic and real-world datasets.