Evaluating machine learning techniques to define the factors related to boar taint

Georgios Makridis, Evert Heyrman, Dimitrios Kotios, Philip Mavrepis, Bert Callens, Ruben Van De Vijver, Jarissa Maselyne, Marijke Aluwé, Dimosthenis Kyriazis

Onderzoeksoutput: Bijdrage aan tijdschriftA1: Web of Science-artikelpeer review



Several industries and sectors such as health care, agriculture, and finance exploit the added value of data to
produce valuable insights for decision-making. The case of so-called ’boar taint’, the unwanted taste and odor
that can be present in meat of entire male pigs, is one real-life scenario that showcases the added value of utilizing
collected data. This information may yield insights for pig farmers about how they could adjust their
management to reduce boar taint. This study examines multiple predictive data-driven approaches coupled with
eXplainable AI (XAI) methods, evaluating them against various explainable metrics while trying to generate
actionable insights and recommendations. Specifically, in this approach, the examined use case was modeled as a
binary classification task resulting in a highly imbalanced dataset. This yielded some functional attributes
regarding the farm/stable and slaughterhouse conditions, such as the type of feed, type of ventilation system,
pharmaceutical treatment, floor type, and the duration of waiting in lairage.
Oorspronkelijke taalEngels
Artikel nummer105045
TijdschriftLivestock Science
Pagina's (van-tot)1-14
Aantal pagina’s14
PublicatiestatusGepubliceerd - okt.-2022


  • B400-veeteelt
  • B420-voeding
  • Boar taint
  • Data analytics
  • Feature importance
  • Imbalanced data
  • Machine learning

Dit citeren