B0531
Title: Aggregating noisy data for improved prediction in multiclass models
Authors: Garth Tarr - University of Sydney (Australia) [presenting]
Ines Wilms - Maastricht University (Netherlands)
Abstract: Faced with changing markets and evolving consumer demands, beef industries are investing in grading systems to maximize value extraction throughout their entire supply chain. The meat standards Australia (MSA) system is a customer-oriented total quality management system that stands out internationally by predicting quality grades of specific muscles processed by a designated cooking method. The model currently underpinning the MSA system requires laborious effort to estimate and its prediction performance may be less accurate in the presence of unbalanced data sets where many muscle-cook combinations have few observations and/or few predictors of palatability are available. Further, the underlying relationships can easily be overwhelmed by the noise inherent in consumer trial data. A novel predictive method is proposed for beef-eating quality that bridges a spectrum of muscle-cook-specific models. At one extreme, each muscle-cook combination is modelled independently; at the other extreme, a pooled predictive model is obtained across all muscle-cook combinations. Via a data-driven regularization method, all muscle-cook-specific models are covered along this spectrum. The proposed predictive method is demonstrated to attain considerable accuracy improvements relative to independent or pooled approaches on unique MSA data sets.
