Adaptive Overlapping Community Detection with Bayesian NonNegative Matrix Factorization

Overlapping Community Detection from a real network is unsupervised, and it is hard to know the exact community number or quantized strength of every node related to each community. Using Non-negative Matrix Factorization (NMF) for Community Detection, we can find two non-negative matrices from whole network adjacent matrix, and the product of two matrices approximates the original matrix well. With Bayesian explanation in factorizing process, we can not only catch most appropriate count of communities in a large network with Shrinkage method, but also verify good threshold how a node should be assigned to a community in fuzzy situation. We apply our approach in some real networks and a synthetic network with benchmark. Experimental results for overlapping community detection show that our method is effective to find the communities number and overlapping degree, and achieve better performance than other existing overlapping community detection methods.