MULTIPLE OUTLIER DETECTION REVISITED

Authors
WALCZAK B MASSART DL
Citation
B. Walczak et Dl. Massart, MULTIPLE OUTLIER DETECTION REVISITED, Chemometrics and intelligent laboratory systems, 41(1), 1998, pp. 1-15
Citations number
28
Categorie Soggetti
Computer Science Artificial Intelligence","Robotics & Automatic Control","Instument & Instrumentation","Chemistry Analytical","Computer Science Artificial Intelligence","Robotics & Automatic Control
Journal title
Chemometrics and intelligent laboratory systems → ACNP
ISSN journal
01697439
Volume
41
Issue
1
Year of publication
1998
Pages
1 - 15
Database
ISI
SICI code
0169-7439(1998)41:1<1:MODR>2.0.ZU;2-I
Abstract
The discrimination power of the classical or/and robust diagnostics fo r the 34 (real and simulated) regression data sets with multiple outli ers is compared. These diagnostics, presented in the uniform way for t he large number of objects, are then used in the pattern recognition a pproaches (PLS, Neural Networks, Rough Set Theory) to estimate the joi nt discrimination power of the classical or/and robust diagnostics and to construct models (or logical rules), allowing identification of ou tliers in the new data sets. (C) 1998 Elsevier Science B.V. All rights reserved.