Collaborative fuzzy clustering algorithm: Some refinements

•Necessity of partition matrices reordering has been examined.•A new collaborative strength optimization method has been given.•Global data structure is formed as a granular partition matrix. Since the inception of the concept of collaborative fuzzy clustering (CFC), many related ideas and algorithm...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:International journal of approximate reasoning Jg. 86; S. 41 - 61
Hauptverfasser: Shen, Yinghua, Pedrycz, Witold
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Inc 01.07.2017
Schlagworte:
Collaborative fuzzy clustering
Collaborative strength optimization
Granular partition matrix
Horizontal mode of clustering
Partition matrix reordering
Partition matrix reordering
Granular partition matrix
Collaborative fuzzy clustering
Horizontal mode of clustering
Collaborative strength optimization
ISSN:0888-613X, 1873-4731
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:•Necessity of partition matrices reordering has been examined.•A new collaborative strength optimization method has been given.•Global data structure is formed as a granular partition matrix. Since the inception of the concept of collaborative fuzzy clustering (CFC), many related ideas and algorithms have been proposed. In this study, we offer a synthetic view of this body of knowledge. We further concentrate on the horizontal version of the CFC algorithm being regarded as one of the major branches of the CFC. Our intent is to address the following three open questions: (a) assessing the necessity of reordering partition matrices prior to invoking the collaboration process; (b) analyzing the impact of linkage strengths on the performance of the clustering results; and (c) forming a representative global data structure with the use of the concept of information granules leading to so-called granular partition matrices. A collection of experimental studies is provided to quantify the underlying concepts and algorithms.
ISSN:0888-613X
1873-4731
DOI:10.1016/j.ijar.2017.04.004