Método heurístico para particionamiento óptimo

Many data analysis problems deal with non supervised partitioning of a data set,in non empty clusters well separated between them and homogeneous within the clusters.An ideal partitioning is obtained when any object can be assigned a class without ambiguity. The present paper has two main parts; fir...

Descripción completa

Autores Principales: de los Cobos Silva, Sergio Gerardo, Trejos Zelaya, Javier, Pérez Salvador, Blanca Rosa, Gutiérrez Andrade, Miguel Ángel
Formato: Artículo
Idioma: Español
Publicado: 2015
Acceso en línea: http://revistas.ucr.ac.cr/index.php/matematica/article/view/221
http://hdl.handle.net/10669/12864
Sumario: Many data analysis problems deal with non supervised partitioning of a data set,in non empty clusters well separated between them and homogeneous within the clusters.An ideal partitioning is obtained when any object can be assigned a class without ambiguity. The present paper has two main parts; first, we present different methodsand heuristics that find the number of clusters for optimal partitioning of a set; afterwards,we propose a new heuristic and we perform different comparisons in orderto evaluate the advantages on well known data sets; we end the paper with someconcluding remarks.Keywords: Optimal partitioning, clustering, classification, heuristics.