Premium
Robust dynamic clustering
Author(s) -
Aboukalam M.A.F.,
AINachawati H.
Publication year - 1992
Publication title -
statistica neerlandica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.52
H-Index - 39
eISSN - 1467-9574
pISSN - 0039-0402
DOI - 10.1111/j.1467-9574.1992.tb01333.x
Subject(s) - cluster analysis , outlier , affine transformation , covariance matrix , algorithm , population , covariance , mathematics , computer science , multivariate normal distribution , multivariate statistics , data mining , mathematical optimization , artificial intelligence , statistics , demography , sociology , pure mathematics
The dynamic clustering (DC) algorithm is a method for discovering clusters in a given population. Unfortunately the classical DC algorithms fail to perform well in the presence of outliers. A robust dynamic clustering (RDC) algorithm is introduced to overcome this problem. Robust estimates of the location vector and the covariance matrix are calculated in the affine invariant case. A simulation study is presented to demonstrate the basic difference between the DC and the RDC algorithms. Three kinds of optimization criteria are used in case of contaminated multivariate normal distributions.