Attributed community mining using joint general non-negative matrix factorization with graph Laplacian
Community mining for complex social networks with link and attribute information plays an important role according to different application needs. In this paper, based on our proposed general non-negative matrix factorization (GNMF) algorithm without dimension matching constraints in our previous work, we propose the joint GNMF with graph Laplacian (LJGNMF) to implement community mining of complex social networks with link and attribute information according to different application needs. Theoretical derivation result shows that the proposed LJGNMF is fully compatible with previous methods of integrating traditional NMF and symmetric NMF. In addition, experimental results show that the proposed LJGNMF can meet the needs of different community minings by adjusting its parameters, and the effect is better than traditional NMF in the community vertices attributes entropy.
Chen, Z., Li, L., Peng, H., Liu, Y., & Yang, Y. (2018). Attributed community mining using joint general non-negative matrix factorization with graph Laplacian. Physica A: Statistical Mechanics and Its Applications, 495, 324–335. https://doi.org/10.1016/j.physa.2017.12.038