CALF-SBM: A covariate-assisted latent factor stochastic block model

Area of Science:

• Network Science
• Statistical Modeling
• Machine Learning

Background:

• Standard stochastic block models (SBMs) do not fully capture network complexity.
• Node-level information and nodal heterogeneity are often overlooked in network analysis.

Purpose of the Study:

• To introduce a novel network generative model, the covariate-assisted latent factor stochastic block model (CALF-SBM).
• To enhance community detection by integrating observed node attributes and addressing network-induced heterogeneity.
• To develop a fully Bayesian inference framework for the proposed model.

Main Methods:

• Extension of the standard stochastic block model.
• Incorporation of node-level covariates and latent factors.
• Fully Bayesian inference framework.
• Model-selection for estimating the number of communities.
• Extensive simulation studies and comparison with existing algorithms.

Main Results:

• The CALF-SBM demonstrates superior performance in community detection compared to classical and modern algorithms.
• The model effectively utilizes node-level information and accounts for nodal heterogeneity.
• Successful application to both simulated and real-world network data.

Conclusions:

• CALF-SBM offers a powerful and flexible framework for network generative modeling and community detection.
• The model provides a robust approach for analyzing complex networks with rich node attributes.
• This method advances the field of network analysis, particularly for applications requiring nuanced community structure identification.