Optimal choice of parameters for a density-based clustering algorithm

Authors:
Wenyan Gan;Deyi Li
Affiliations:
Nanjing University of Science and Technology, Nanjing, China;Institute of Electronic System Engineering, Beijing, China
Venue:
RSFDGrC'03 Proceedings of the 9th international conference on Rough sets, fuzzy sets, data mining, and granular computing
Year:
2003

Citing 5
Cited 1

Scale-Space for Discrete Signals

IEEE Transactions on Pattern Analysis and Machine Intelligence
Information theory

Information theory
The Choice of Origin and Scale for Graphs

Journal of the ACM (JACM)
Clustering by Scale-Space Filtering

IEEE Transactions on Pattern Analysis and Machine Intelligence
Data Mining: Concepts and Techniques

Data Mining: Concepts and Techniques

A non-parametric heuristic algorithm for convex and non-convex data clustering based on equipotential surfaces

Expert Systems with Applications: An International Journal

Quantified Score

Hi-index	0.00

Visualization

Abstract

Clustering is an important and challenging task in data mining. As a kind of generalized density-based clustering methods, DENCLUE algorithm has many remarkable properties, but the quality of clustering results strongly depends on the adequate choice of two parameters: density parameter σ and noise threshold ξ. In this paper, by investigating the influence of the two parameters of DENCLUE algorithm on the clustering results, we firstly show that an optimal σ should be chosen to obtain good clustering results. Then, an entropy-based method is proposed for the optimal choice of σ. Further, noise threshold ξ is estimated to produce a reasonable pattern of clustering. Finally, experiments are performed to illustrate the effectiveness of our methods.