A classification model for three-way dissimilarity data is presented including the idea to identify simultaneously a fuzzy partition of the set of occasions in classes of similar proximity matrices and, within each class, to fit a 2-ultrametric matrix (i.e., with at most two different off-diagonal values) that represents the best consensus partition of the objects. We propose a data reduction approach that induces a specific covariance structure inside each class of the fuzzy partition. The model is formalized by a constrained maximum likelihood mixed-integer problem and a numerical method, based on ECM strategy, is also described.
A classification model for three-way dissimilarity data is presented including the idea to identify simultaneously a fuzzy partition of the set of occasions in classes of similar proximity matrices and, within each class, to fit a 2-ultrametric matrix (i.e., with at most two different off-diagonal values) that represents the best consensus partition of the objects. We propose a data reduction approach that induces a specific covariance structure inside each class of the fuzzy partition. The model is formalized by a constrained maximum likelihood mixed-integer problem and a numerical method, based on ECM strategy, is also described.
Classification of Three-way Proximity Data / Bocci, Laura; Vicari, Donatella; Vichi, Maurizio. - STAMPA. - (2002), pp. 40-40. (Intervento presentato al convegno IFCS - Data Analysis, Classification, and Related Methods tenutosi a Cracow (Poland) nel July 16-19, 2002).
Classification of Three-way Proximity Data
BOCCI, Laura;VICARI, Donatella;VICHI, Maurizio
2002
Abstract
A classification model for three-way dissimilarity data is presented including the idea to identify simultaneously a fuzzy partition of the set of occasions in classes of similar proximity matrices and, within each class, to fit a 2-ultrametric matrix (i.e., with at most two different off-diagonal values) that represents the best consensus partition of the objects. We propose a data reduction approach that induces a specific covariance structure inside each class of the fuzzy partition. The model is formalized by a constrained maximum likelihood mixed-integer problem and a numerical method, based on ECM strategy, is also described.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.