Improved multi-objective clustering algorithm using particle swarm optimization

Multi-objective clustering has received widespread attention recently, as it can obtain more accurate and reasonable solution. In this paper, an improved multi-objective clustering framework using particle swarm optimization (IMCPSO) is proposed. Firstly, a novel particle representation for clusteri...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:PloS one Jg. 12; H. 12; S. e0188815
Hauptverfasser: Gong, Congcong, Chen, Haisong, He, Weixiong, Zhang, Zhanliang
Format: Journal Article
Sprache:Englisch
Veröffentlicht: United States Public Library of Science 05.12.2017
Public Library of Science (PLoS)
Schlagworte:
Algorithms
Analysis
Biology and Life Sciences
Cluster analysis
Clustering
Computer and Information Sciences
Data analysis
Datasets
Earth Sciences
Fuzzy sets
Information science
Intelligence
International conferences
Medicine and Health Sciences
Methods
Multiple objective analysis
Optimization
Optimization theory
Particle swarm optimization
Physical Sciences
Problems
Research and Analysis Methods
Researchers
Social Sciences
Studies
Swarm intelligence
ISSN:1932-6203, 1932-6203
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Multi-objective clustering has received widespread attention recently, as it can obtain more accurate and reasonable solution. In this paper, an improved multi-objective clustering framework using particle swarm optimization (IMCPSO) is proposed. Firstly, a novel particle representation for clustering problem is designed to help PSO search clustering solutions in continuous space. Secondly, the distribution of Pareto set is analyzed. The analysis results are applied to the leader selection strategy, and make algorithm avoid trapping in local optimum. Moreover, a clustering solution-improved method is proposed, which can increase the efficiency in searching clustering solution greatly. In the experiments, 28 datasets are used and nine state-of-the-art clustering algorithms are compared, the proposed method is superior to other approaches in the evaluation index ARI.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
Competing Interests: The authors have declared that no competing interests exist.
ISSN:1932-6203
1932-6203
DOI:10.1371/journal.pone.0188815