Decision tree, a learning tool for the prediction of beef tenderness using rearing factors and carcass characteristics

Authors:
Publisher website
Google Scholar
Abstract
BACKGROUND: The present study explored the potential use of decision trees on rearing factors (q = 10) and carcass characteristics (q = 12) for the development of prediction model rules of beef tenderness prediction/categorization. Accordingly, 308 young bulls were used by a sensory panel to evaluate the tenderness potential of ribeye steaks grilled at 55 °C. A classification and regression tree method was implemented and allowed the prediction of tenderness using (i) rearing factors, (ii) carcass characteristics or (iii) both.RESULTS: The resultant tree models yielded predictive accuracies of 70.78% (with four rearing factors: concentrate percentage; fattening duration; initial body weight and dry matter intake); 67.21% (with four carcass characteristics: fatness carcass score; carcass weight; dressing percentage and muscle carcass percentage) and 84.41% (with six rearing factors and carcass characteristics) compared to the k‐means clustering of tenderness. In the final and robust regression tree, from the 22 attribute information, two carcass characteristics (fatness carcass score and muscle carcass percentage) and four rearing factors (fattening duration; concentrate percentage; dry matter intake and initial body weight) were retained as predictors. The first splitter of the 308 ribeye steaks in accordance with their tenderness scores was fatness carcass score, followed by fattening duration and concentrate percentage.CONCLUSION: The trial in the preset study highlights the importance of thresholding approach for efficiently classifying ribeye steaks in accordance with their tenderness potential. The overall prediction model rule was: IF (fatness carcass score ≥ 2.88) AND (concentrate ≥ 82%) [AND (muscle carcass ≥ 71%)] THEN meat was [very] tender. © 2018 Society of Chemical Industry
Funding Information
  • Fonds Européens de Développement Régional
  • Région Auvergne-Rhône-Alpes

This publication has 40 references indexed in Scilit: