V. Makarenkov et P. Legendre, Optimal variable weighting for ultrametric and additive trees and K-means partitioning: Methods and software, J CLASSIF, 18(2), 2001, pp. 245-271
De Soete (1986, 1988) proposed some years ago a method for optimal variable
weighting for ultrametric. and additive tree fitting. This paper extends D
e Soete's method to optimal variable weighting for K-means partitioning. We
also describe some new features and improvements to the algorithm proposed
by De Soete. Monte Carlo simulations have been conducted using different e
rror conditions. In all cases (i.e., ultrametric or additive trees, or K-me
ans partitioning), the simulation results indicate that the optimal weighti
ng procedure should be used for analyzing data containing noisy variables t
hat do not contribute relevant information to the classification structure.
However, if the data involve error-perturbed variables that are relevant t
o the classification or outliers, it seems better to cluster or partition t
he entities by using variables with equal weights, A new computer progam, O
VW, which is available to researchers as freeware, implements improved algo
rithms for optimal variable weighting for ultrametric and additive tree clu
stering, and includes a new algorithm for optimal variable weighting for K-
means partitioning.