Using partial least squares regression and multiplicative scatter correction for FT-NIR data evaluation of wheat flours

Jaana Sorvaniemi, Arvo Kinnunen, Andrew Tsados, Yrjö Mälkki

Research output: Contribution to journalArticleScientificpeer-review

16 Citations (Scopus)

Abstract

Scanning Fourier transform near infrared reflectance (FT-NIR) spectroscopy was used to record spectra of roller and commercial milled wheat flours (varieties and mixtures). Partial least squares (PLS) regression of multiplicative scatter corrected (MSC) NIR data was carried out. Correlations to moisture, protein, wet gluten, water absorption and falling number were studied. Promising correlations to these parameters except falling number were obtained for both types of milled samples. Explanation factors (R2) and standard errors of prediction (SEP) in the roller milled flours were as follows: moisture R2 = 0.97 and SEP = O.16 g/100 g; protein R2 = 0.92 and SEP = 0.40 g/100 g d.b. (dry basis); wet gluten R2 = 0.89 and SEP = 1.37 g/100 g; and water absorption R2 = 0.83 and SEP = 0.84 g/100 g. The explanation factor of predicted falling numbers to the actual values of the roller milled flours remained lower than 0.5. Explanation factors and SEP in the commercial milled flours were as follows: moisture R2 = 0.85 and SEP = 0.19 g/100 g; protein R2 = 0.91 and SEP = 0.63 g/100 g d.b.; wet gluten R2 = 0.87 and SEP = 2.22 g/100 g; water absorption R2 = 0.88 and SEP = 1.10 g/100 g; and falling number R2 = 0.55 and SEP = 28.2.

Original languageEnglish
Pages (from-to)251 - 258
Number of pages8
JournalLWT - Food Science and Technology
Volume26
Issue number3
DOIs
Publication statusPublished - 1993
MoE publication typeA1 Journal article-refereed

Fingerprint Dive into the research topics of 'Using partial least squares regression and multiplicative scatter correction for FT-NIR data evaluation of wheat flours'. Together they form a unique fingerprint.

  • Cite this