Abstract
The performance of an iterative algorithm for finding the most suitable regression data to be used for the regression of local quantitative partial least-squares (PLS) models was investigated. This algorithm represented a further development of a method called multilayer PLS (ML-PLS) in previous publications, including automated distance measure selection and evaluation of how many iterations are to be used. The investigation was carried out by creating ML-PLS models based on data previously published and made available in different ways. The quantitative applications of near infrared spectroscopy that were studied were measurements of clay in soil samples, protein content in single kernels, freezing points in diesel fuels and ammonium in spiked anaerobic digestion samples. With these data sets, 20–50% improvements in the measurement accuracy compared with global PLS models were obtained. These accuracies obtained with ML-PLS were also compared with accuracies previously reported in the literature for the same data. Also, based on this comparison, the results obtained using the ML-PLS algorithm were encouraging.
© 2014 IM Publications LLP
PDF Article
More Like This
Cited By
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
Contact your librarian or system administrator
or
Login to access Optica Member Subscription