Conceptual clustering of multi-relational data

• Type: Article in International Conference Proceedings Book
• Date: 2012

Authors

Fonseca, NA (Author) FCUP; Santos Costa, V (Author) FCUP; Camacho, R (Author) FEUP

Conference proceedings International

Title: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Pages: 145-159

21st International Conference on InductiveLogic Programming, ILP 2011

Windsor Great Park, 31 July 2011 through 3 August 2012

Indexing

Scopus - 2 Citations

Other information

Authenticus ID: P-008-5F3

DOI: 10.1007/978-3-642-31951-8_16

Abstract (EN): "Traditional" clustering, in broad sense, aims at organizing objects into groups (clusters) whose members are "similar" among them and are "dissimilar" to objects belonging to the other groups. In contrast, in conceptual clustering the underlying structure of the data together with the description language which is available to the learner is what drives cluster formation, thus providing intelligible descriptions of the clusters, facilitating their interpretation. We present a novel conceptual clustering system for multi-relational data, based on the popular k¿-¿medoids algorithm. Although clustering is, generally, not straightforward to evaluate, experimental results on several applications show promising results. Clusters generated without class information agree very well with the true class labels of cluster's members. Moreover, it was possible to obtain intelligible and meaningful descriptions of the clusters. © 2012 Springer-Verlag Berlin Heidelberg.

Language: English

Type (Professor's evaluation): Scientific