In food authenticity studies the central concern is the detection of prod- ucts that are not what they claim to be. Here, we introduce robustness in a semi- supervised classification rule, to identify non-authentic sub-samples. The approach is based on discriminating observations with the lowest contributions to the over- all likelihood, following the impartial trimming established technique. Experiments on real data, artificially adulterated, are provided to underline the benefits of the proposed method.
Negli studi di autenticit`a degli alimenti risulta cruciale saper riconoscere prodotti contraffatti. In questo paper si adotta un approccio robusto per modificare una regola di classificazione semi-supervised e poter quindi identificare potenziali adulterazioni. L’approccio basato sulla selezione delle osservazioni che danno mi- nore contributo alla verosimiglianza globale, seguendo tecniche ben note di im- partial trimming. Esperimenti su dati reali, artificialmente adulterati, evidenziano l’efficacia del metodo proposto.
Cappozzo, A., Greselin, F., Murphy, B., Robust Updating Classification Rule with applications in Food Authenticity Studies = Robust Updating Classification Rule con applicazioni a studi di autenticità degli alimenti, Comunicazione, in Book of short Papers SIS 2018, (Palermo, 20-22 June 2018), Pearson, Palermo 2018: 1-6 [https://hdl.handle.net/10807/306442]
Robust Updating Classification Rule with applications in Food Authenticity Studies = Robust Updating Classification Rule con applicazioni a studi di autenticità degli alimenti
Cappozzo, Andrea;
2018
Abstract
In food authenticity studies the central concern is the detection of prod- ucts that are not what they claim to be. Here, we introduce robustness in a semi- supervised classification rule, to identify non-authentic sub-samples. The approach is based on discriminating observations with the lowest contributions to the over- all likelihood, following the impartial trimming established technique. Experiments on real data, artificially adulterated, are provided to underline the benefits of the proposed method.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.