Taylor & Francis Group
Browse
1/1
2 files

Romanian bee pollen classification and property modelling

dataset
posted on 2020-01-23, 15:45 authored by Raluca Daniela Isopescu, Roxana Spulber, Ana Maria Josceanu, Dan Eduard Mihaiescu, Ovidiu Popa

The 4000–400 cm−1 spectral information obtained by FTIR-ATR for bee pollen samples of different floral origin from various Romanian counties were used to formulate a method for rapid classification using principal component analysis (PCA) and linear discriminant analysis (LDA) as multivariate statistics tools. Entire IR range spectra were used as input data in both techniques. PCA reduced the problem dimensionality and put in evidence the presence of outliers. LDA, applied for samples characterized by the first 11 principal components (PCA-LDA), clearly separated 10 groups according to botanical and geographical origin, thus representing a base for developing an identification tool of the new pollen samples origin. Furthermore, spectral information and previously determined properties (polyphenols, flavonoids, sugars, proteins) were tested for identifying polynomial correlations using partial least squares (PLS) technique. The good correlation coefficients, R2 > 0.97, and low mean errors in both training and testing stages proved that a FTIR-ATR spectrum provides key information for rapid screening of bee pollen in terms of the parameters significant for its nutritional labelling.

History

Usage metrics

    Journal of Apicultural Research

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC