Premium
Estimation of an Optimized Number of Topics by Consensus Soft Clustering Using NMF
Author(s) -
Yokoi Takeru
Publication year - 2013
Publication title -
electronics and communications in japan
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.131
H-Index - 13
eISSN - 1942-9541
pISSN - 1942-9533
DOI - 10.1002/ecj.11438
Subject(s) - cluster analysis , non negative matrix factorization , computer science , set (abstract data type) , heuristic , data mining , document clustering , correlation clustering , consensus clustering , artificial intelligence , matrix decomposition , pattern recognition (psychology) , cure data clustering algorithm , eigenvalues and eigenvectors , physics , quantum mechanics , programming language
SUMMARY We propose here a novel approach to exploring an optimized number of topics in a document set using consensus clustering based on nonnegative matrix factorization ( NMF ). It is useful to automatically determine the number of topics in a document set because various approaches to heuristic topic extraction determine it. Consensus clustering merges multiple results of clustering so as to achieve robust clustering. In this paper, assuming that robust clustering is achieved by optimizing the number of clusters, we propose a novel consensus soft clustering algorithm based on NMF and estimate an optimized number of topics with exploration of a robust classification of documents into topics. © 2013 Wiley Periodicals, Inc. Electron Comm Jpn, 96(8): 50–58, 2013; Published online in Wiley Online Library ( wileyonlinelibrary.com ). DOI 10.1002/ecj.11438