Community Extraction in Multilayer Networks with Heterogeneous Community Structure

Area of Science:

• Network Science
• Data Mining
• Graph Theory

Background:

• Multilayer networks model complex relationships, but community detection methods are underdeveloped.
• Existing methods struggle with heterogeneous layers and overlapping communities.

Purpose of the Study:

• To introduce and evaluate Multilayer Extraction, a new procedure for community detection in multilayer networks.
• To address limitations in current methods for analyzing complex network structures.

Main Methods:

• Multilayer Extraction uses a significance-based score comparing observed connectivity to a random graph model.
• The procedure handles heterogeneous layers and can identify overlapping communities and background elements.
• Theoretical consistency is established under the multilayer stochastic block model.

Main Results:

• The method successfully identifies densely connected vertex-layer sets in multilayer networks.
• Evaluations on applications and simulations demonstrate its effectiveness.
• It captures overlapping communities and distinguishes them from background noise.

Conclusions:

• Multilayer Extraction is a robust and effective exploratory tool for analyzing complex multilayer networks.
• The method offers advancements in understanding multi-relational data structures.
• Publicly available code facilitates its application in network analysis.