Classification of cow diet based on milk Mid Infrared Spectra: A data analysis competition at the “International Workshop on Spectroscopy and Chemometrics 2022”

What is to be gained by ensemble models in analysis of spectroscopic data?

“…The dataset comprised 3,275 milk MIR spectra of animals fed one of diet regiments: grass-only (GRS), grass-white clover (CLV) and nutional mix containing grass silage, maize silage, and concentrates (TMR) O' Callaghan et al (2016); Frizzarin et al (2021cFrizzarin et al ( , 2023. The task was to predict animal diet type from MIR spectral measurements of the milk samples.…”

“…Datasets used in this study originated from two data analysis challenges organised by the Vistamilk SFI Research Centre as part of International Workshop on Spectroscopy and Chemometrics 2021 Frizzarin et al (2021aFrizzarin et al (2023. Both competitions involved a supervised learning task from mid-infrared (MIR) spectra of cow milk samples.…”

“…For the classification task, Fisher score (ratio of between to within diet group variance) and genetic algorithm (GA) Holland (2019) were used to further reduce the number of input wavelengths. These methods are implemented in R libraries Domijan (2018) and Willighagen and Ballings (2022) respectively and the workflow is described in Frizzarin et al (2023).…”

“…In this article, an empirical study was carried out with the aim of evaluating to what extent ensemble models can improve prediction from individual candidate models for spectroscopic data. For this purpose, datasets from two chemometric data analysis competitions Frizzarin et al (2021aFrizzarin et al ( , 2023 were used. The two challenges were to build better calibration models for prediction of milk quality traits and animal diet from mid-infrared (MIR) spectra.…”

“…Many different statistical modelling approaches have been employed for challenges in Frizzarin et al (2021a) and Frizzarin et al (2023). The papers outline different pre-processing steps taken by the participants including outlier removal, variable selection and variable transformation.…”