Selection of the number of clusters via the bootstrap method

Authors:
Yixin Fang;Junhui Wang
Affiliations:
Division of Biostatistics, Department of Environmental Medicine, New York University, 650 First Avenue, Room 551, New York, NY 10016, United States;Department of Mathematics, Statistics, and Computer Science, University of Illinois at Chicago, United States
Venue:
Computational Statistics & Data Analysis
Year:
2012

Citing 4
Cited 2

Normalized Cuts and Image Segmentation

IEEE Transactions on Pattern Analysis and Machine Intelligence
Stability-based validation of clustering solutions

Neural Computation
Penalized cluster analysis with applications to family data

Computational Statistics & Data Analysis
A sober look at clustering stability

COLT'06 Proceedings of the 19th annual conference on Learning Theory

A non-parametric method to estimate the number of clusters

Computational Statistics & Data Analysis
Vector field k-means: clustering trajectories by fitting multiple vector fields

EuroVis '13 Proceedings of the 15th Eurographics Conference on Visualization

Quantified Score

Hi-index	0.03

Visualization

Abstract

Here the problem of selecting the number of clusters in cluster analysis is considered. Recently, the concept of clustering stability, which measures the robustness of any given clustering algorithm, has been utilized in Wang (2010) for selecting the number of clusters through cross validation. In this paper, an estimation scheme for clustering instability is developed based on the bootstrap, and then the number of clusters is selected so that the corresponding estimated clustering instability is minimized. The proposed selection criterion's effectiveness is demonstrated on simulations and real examples.