Cluster Analysis consists of the aggregation of data items of a given set into subsets based on some similarity properties. Clustering techniques have been applied in many fields which typically involve a large amount of complex data. This study focuses on what we call multi-domain clustering and labeling, i.e. a set of techniques for multi-dimensional structured mixed data clustering. The work consists of studying the best mix of clustering techniques that address the problem in the multi-domain setting. Considered data types are numerical, categorical and textual. All of them can appear together within the same clustering scenario. We focus on k-means and agglomerative hierarchical clustering methods based on a new distance function we define for this specific setting. The proposed approach has been validated on some real and realistic data-sets based onto college, automobile and leisure fields. Experimental data allowed to evaluate the effectiveness of the different solutions, both for clustering and labeling.

Clustering and Labeling of Multi-dimensional Mixed Structured Data

BRAMBILLA, MARCO;ZANONI, MASSIMILIANO
2012

Abstract

Cluster Analysis consists of the aggregation of data items of a given set into subsets based on some similarity properties. Clustering techniques have been applied in many fields which typically involve a large amount of complex data. This study focuses on what we call multi-domain clustering and labeling, i.e. a set of techniques for multi-dimensional structured mixed data clustering. The work consists of studying the best mix of clustering techniques that address the problem in the multi-domain setting. Considered data types are numerical, categorical and textual. All of them can appear together within the same clustering scenario. We focus on k-means and agglomerative hierarchical clustering methods based on a new distance function we define for this specific setting. The proposed approach has been validated on some real and realistic data-sets based onto college, automobile and leisure fields. Experimental data allowed to evaluate the effectiveness of the different solutions, both for clustering and labeling.
Search Computing - Broadening Web Search
9783642342127
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/691483
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? ND
social impact