Mid infrared spectroscopy and milk quality traits: a data analysis competition at the "International Workshop on Spectroscopy and Chemometrics 2021"

TL;DR

This paper discusses a data analysis competition focused on predicting milk quality traits from mid-infrared spectra, highlighting various modeling approaches and insights gained from participant solutions.

Contribution

It presents a comparative analysis of different chemometric methods used in a milk quality prediction challenge based on spectral data.

Findings

Different modeling strategies were employed by participants.

Insights into trait-specific modeling challenges.

Critical discussion of approaches and results.

Abstract

A chemometric data analysis challenge has been arranged during the first edition of the "International Workshop on Spectroscopy and Chemometrics", organized by the Vistamilk SFI Research Centre and held online in April 2021. The aim of the competition was to build a calibration model in order to predict milk quality traits exploiting the information contained in mid-infrared spectra only. Three different traits have been provided, presenting heterogeneous degrees of prediction complexity thus possibly requiring trait-specific modelling choices. In this paper the different approaches adopted by the participants are outlined and the insights obtained from the analyses are critically discussed.