Community Detection via Multihop Nonnegative Matrix Factorization.

Area of Science:

• Graph theory
• Network analysis
• Data mining

Background:

• Community detection is crucial for analyzing complex networks in various fields.
• Nonnegative matrix factorization (NMF) is a popular approach for community detection.
• Existing NMF methods often overlook multihop connectivity, limiting their effectiveness.

Purpose of the Study:

• To propose a novel community detection method that incorporates multihop connectivity patterns.
• To develop an efficient optimization algorithm for the proposed method.
• To evaluate the performance of the new method against state-of-the-art techniques.

Main Methods:

• Development of multihop nonnegative matrix factorization (MHNMF).
• Derivation of an efficient algorithm for MHNMF optimization.
• Theoretical analysis of computational complexity and convergence.
• Experimental validation on 12 real-world benchmark networks.

Main Results:

• MHNMF successfully integrates multihop connectivity information into community detection.
• The proposed optimization algorithm is efficient and converges theoretically.
• MHNMF demonstrates superior performance compared to 12 existing community detection methods on benchmark datasets.

Conclusions:

• MHNMF offers an effective approach for community detection by leveraging multihop network structures.
• The method provides a significant advancement over traditional NMF-based techniques.
• MHNMF shows strong potential for applications requiring accurate community identification in complex networks.