Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data
In multivariate calibrations, locally weighted partial least squared regression (LWPLSR) is an efficient prediction method when heterogeneity of data generates nonlinear relations (curvatures and clustering) between the response and the explicative variables. This is frequent in agronomic data sets that gather materials of different natures or origins. LWPLSR is a particular case of weighted PLSR (WPLSR; ie, a statistical weight different from the standard 1/n is given to each of the n calibration observations for calculating the PLS scores/loadings and the predictions). In LWPLSR, the weights depend from the dissimilarity (which has to be defined and calculated) to the new observation to predict. This article compares two strategies of LWPLSR: (a) “LW”: the usual strategy where, for each new observation to predict, a WPLSR is applied to the n calibration observations (ie, entire calibration set) vs (b) “KNN‐LW”: a number of k nearest neighbors to the observation to predict are preliminary selected in the training set and WPLSR is applied only to this selected KNN set. On three illustrating agronomic data sets (quantitative and discrimination predictions), both strategies overpassed the standard PLSR. LW and KNN‐LW had close prediction performances, but KNN‐LW was much faster in computation time. KNN‐LW strategy is therefore recommended for large data sets. The article also presents a new algorithm for WPLSR, on the basis of the “improved kernel #1” algorithm, which is competitor and in general faster to the already published weighted PLS nonlinear iterative partial least squares (NIPALS).
Main Authors: | , , |
---|---|
Format: | article biblioteca |
Language: | eng |
Subjects: | U10 - Informatique, mathématiques et statistiques, U30 - Méthodes de recherche, A50 - Recherche agronomique, analyse de données, spectroscopie infrarouge, http://aims.fao.org/aos/agrovoc/c_15962, http://aims.fao.org/aos/agrovoc/c_28568, |
Online Access: | http://agritrop.cirad.fr/595227/ http://agritrop.cirad.fr/595227/7/595227.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
id |
dig-cirad-fr-595227 |
---|---|
record_format |
koha |
spelling |
dig-cirad-fr-5952272024-01-29T02:39:21Z http://agritrop.cirad.fr/595227/ http://agritrop.cirad.fr/595227/ Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data. Lesnoff Matthieu, Metz Maxime, Roger Jean-Michel. 2020. Journal of Chemometrics, 34 (5):e3209, 13 p.https://doi.org/10.1002/cem.3209 <https://doi.org/10.1002/cem.3209> Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data Lesnoff, Matthieu Metz, Maxime Roger, Jean-Michel eng 2020 Journal of Chemometrics U10 - Informatique, mathématiques et statistiques U30 - Méthodes de recherche A50 - Recherche agronomique analyse de données spectroscopie infrarouge http://aims.fao.org/aos/agrovoc/c_15962 http://aims.fao.org/aos/agrovoc/c_28568 In multivariate calibrations, locally weighted partial least squared regression (LWPLSR) is an efficient prediction method when heterogeneity of data generates nonlinear relations (curvatures and clustering) between the response and the explicative variables. This is frequent in agronomic data sets that gather materials of different natures or origins. LWPLSR is a particular case of weighted PLSR (WPLSR; ie, a statistical weight different from the standard 1/n is given to each of the n calibration observations for calculating the PLS scores/loadings and the predictions). In LWPLSR, the weights depend from the dissimilarity (which has to be defined and calculated) to the new observation to predict. This article compares two strategies of LWPLSR: (a) “LW”: the usual strategy where, for each new observation to predict, a WPLSR is applied to the n calibration observations (ie, entire calibration set) vs (b) “KNN‐LW”: a number of k nearest neighbors to the observation to predict are preliminary selected in the training set and WPLSR is applied only to this selected KNN set. On three illustrating agronomic data sets (quantitative and discrimination predictions), both strategies overpassed the standard PLSR. LW and KNN‐LW had close prediction performances, but KNN‐LW was much faster in computation time. KNN‐LW strategy is therefore recommended for large data sets. The article also presents a new algorithm for WPLSR, on the basis of the “improved kernel #1” algorithm, which is competitor and in general faster to the already published weighted PLS nonlinear iterative partial least squares (NIPALS). article info:eu-repo/semantics/article Journal Article info:eu-repo/semantics/publishedVersion http://agritrop.cirad.fr/595227/7/595227.pdf text Cirad license info:eu-repo/semantics/restrictedAccess https://agritrop.cirad.fr/mention_legale.html https://doi.org/10.1002/cem.3209 10.1002/cem.3209 info:eu-repo/semantics/altIdentifier/doi/10.1002/cem.3209 info:eu-repo/semantics/altIdentifier/purl/https://doi.org/10.1002/cem.3209 |
institution |
CIRAD FR |
collection |
DSpace |
country |
Francia |
countrycode |
FR |
component |
Bibliográfico |
access |
En linea |
databasecode |
dig-cirad-fr |
tag |
biblioteca |
region |
Europa del Oeste |
libraryname |
Biblioteca del CIRAD Francia |
language |
eng |
topic |
U10 - Informatique, mathématiques et statistiques U30 - Méthodes de recherche A50 - Recherche agronomique analyse de données spectroscopie infrarouge http://aims.fao.org/aos/agrovoc/c_15962 http://aims.fao.org/aos/agrovoc/c_28568 U10 - Informatique, mathématiques et statistiques U30 - Méthodes de recherche A50 - Recherche agronomique analyse de données spectroscopie infrarouge http://aims.fao.org/aos/agrovoc/c_15962 http://aims.fao.org/aos/agrovoc/c_28568 |
spellingShingle |
U10 - Informatique, mathématiques et statistiques U30 - Méthodes de recherche A50 - Recherche agronomique analyse de données spectroscopie infrarouge http://aims.fao.org/aos/agrovoc/c_15962 http://aims.fao.org/aos/agrovoc/c_28568 U10 - Informatique, mathématiques et statistiques U30 - Méthodes de recherche A50 - Recherche agronomique analyse de données spectroscopie infrarouge http://aims.fao.org/aos/agrovoc/c_15962 http://aims.fao.org/aos/agrovoc/c_28568 Lesnoff, Matthieu Metz, Maxime Roger, Jean-Michel Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data |
description |
In multivariate calibrations, locally weighted partial least squared regression (LWPLSR) is an efficient prediction method when heterogeneity of data generates nonlinear relations (curvatures and clustering) between the response and the explicative variables. This is frequent in agronomic data sets that gather materials of different natures or origins. LWPLSR is a particular case of weighted PLSR (WPLSR; ie, a statistical weight different from the standard 1/n is given to each of the n calibration observations for calculating the PLS scores/loadings and the predictions). In LWPLSR, the weights depend from the dissimilarity (which has to be defined and calculated) to the new observation to predict. This article compares two strategies of LWPLSR: (a) “LW”: the usual strategy where, for each new observation to predict, a WPLSR is applied to the n calibration observations (ie, entire calibration set) vs (b) “KNN‐LW”: a number of k nearest neighbors to the observation to predict are preliminary selected in the training set and WPLSR is applied only to this selected KNN set. On three illustrating agronomic data sets (quantitative and discrimination predictions), both strategies overpassed the standard PLSR. LW and KNN‐LW had close prediction performances, but KNN‐LW was much faster in computation time. KNN‐LW strategy is therefore recommended for large data sets. The article also presents a new algorithm for WPLSR, on the basis of the “improved kernel #1” algorithm, which is competitor and in general faster to the already published weighted PLS nonlinear iterative partial least squares (NIPALS). |
format |
article |
topic_facet |
U10 - Informatique, mathématiques et statistiques U30 - Méthodes de recherche A50 - Recherche agronomique analyse de données spectroscopie infrarouge http://aims.fao.org/aos/agrovoc/c_15962 http://aims.fao.org/aos/agrovoc/c_28568 |
author |
Lesnoff, Matthieu Metz, Maxime Roger, Jean-Michel |
author_facet |
Lesnoff, Matthieu Metz, Maxime Roger, Jean-Michel |
author_sort |
Lesnoff, Matthieu |
title |
Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data |
title_short |
Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data |
title_full |
Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data |
title_fullStr |
Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data |
title_full_unstemmed |
Comparison of locally weighted PLS strategies for regression and discrimination on agronomic NIR data |
title_sort |
comparison of locally weighted pls strategies for regression and discrimination on agronomic nir data |
url |
http://agritrop.cirad.fr/595227/ http://agritrop.cirad.fr/595227/7/595227.pdf |
work_keys_str_mv |
AT lesnoffmatthieu comparisonoflocallyweightedplsstrategiesforregressionanddiscriminationonagronomicnirdata AT metzmaxime comparisonoflocallyweightedplsstrategiesforregressionanddiscriminationonagronomicnirdata AT rogerjeanmichel comparisonoflocallyweightedplsstrategiesforregressionanddiscriminationonagronomicnirdata |
_version_ |
1792499918512848896 |