Orthogonal Signal Correction Inside Cross-validation Improves Error Estimation in PLS Models

N. Andersson and D.I. Wilson (Sweden)

Keywords

PLS; cross-validation; Pre-processing; Near Infra Red Spectroscopy; OSC

Abstract

The importance of applying pre-processing filters inside the validation procedure during PLS model building, in stead of applied on the entire data set, is demonstrated us ing NIR spectral data from several applications and an industrial case study. Incorrectly applying the pre-processing will underestimate the model uncertainty, potentially de laying the signalling of process disturbances when used for online plant diagnostics or multivariate statistical process control.

Important Links:



Go Back