Premium
Parsimonious multiscale classification models
Author(s) -
Alsberg Bjørn K.
Publication year - 2000
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/1099-128x(200009/12)14:5/6<529::aid-cem629>3.0.co;2-e
Subject(s) - wavelet , pattern recognition (psychology) , computer science , scale (ratio) , artificial intelligence , feature selection , variable (mathematics) , wavelet transform , partial least squares regression , model selection , mathematics , measure (data warehouse) , multivariate statistics , interpretation (philosophy) , linear discriminant analysis , discriminant , algorithm , data mining , machine learning , programming language , mathematical analysis , physics , quantum mechanics
Wavelet transforms can be used to construct parsimonious multivariate classification models of spectral data. The basic idea is to use a combination of both wavelet data compression and variable/scale selection. The simplest approach determines the optimal resolution level with respect to prediction error and a model complexity measure. A classification model that maintains an acceptable prediction error using only scales with relatively low frequency content can be said to be parsimonious. In addition, it is possible to enhance the interpretation of the classification model by identifying broad or narrow features in the spectral profiles that are important for the prediction. However, for the simplest approach it is more difficult to determine the wavelength domain localization of these features compared to wavelength‐scale variable selection methods. The discriminant partial least squares (DPLS) method is used to demonstrate the feasibility of these approaches to parsimonious model building. Copyright © 2000 John Wiley & Sons, Ltd.