A Theoretical Framework for Supporting Clustering Validation via Non-Negative-Matrix-Factorization Trace Sequences Over Probabilistic Spaces

This paper focuses the attention on the annoying problem of clustering validation, intended as a relevant task of major analytical processing. The proposed method argues to build and analyze a data matrix from probabilistic spaces and to exploit Non-negative Matrix Factorization (NMF) to identify existing clusters and select them according to a Bayesian inter-pretation.