Machine learning-based classification of Kangal Akkaraman and Kivircik lamb carcasses using carcass traits, meat quality, and fatty acid profiles
Small Ruminant Research, cilt.264, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 264
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.smallrumres.2026.107892
- Dergi Adı: Small Ruminant Research
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, BIOSIS, Academic Search Ultimate (EBSCO)
- Anahtar Kelimeler: Breed discrimination, Carcass traits, Linear discriminant analysis, Machine learning, Support vector machines
- Sivas Cumhuriyet Üniversitesi Adresli: Evet
Özet
The aim was to evaluate the classifiability of Kangal Akkaraman and Kivircik breed lamb carcasses using machine learning (ML) methods based on carcass characteristics, carcass dimensions, meat quality, and fatty acid composition data. Data from a total of 155 lamb carcasses, including 60 Kangal Akkaraman and 95 Kivircik breeds, were used. Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), K-Nearest Neighbors (KNN), Support Vector Machines (SVM), Random Forest (RF), XGBoost, and Decision Tree (DT) algorithms were applied. In Dataset-1, where carcass conformation and fatness measures, percentages of carcass joints, and fat colour parameters are predictors, all carcasses could be accurately classified using SVM algorithm with LOOCV validation. In Dataset-2, where carcass dimensions are the predictors, classification was performed with LDA, QDA, SVM, and XGBoost algorithms with an accuracy of 0.994. Classification was performed with Dataset-3, where meat quality characteristics were the predictor, with an accuracy of 0.942, and with Dataset-4, where percentages of individual fatty acids were the predictor, with an accuracy of 0.974. According to LDA results, the most important features in breed differentiation were fatness and conformation in Dataset-1; hind limb length and chest width in Dataset-2; meat yellowness and cooking loss in Dataset-3; and C22:0 and C15:1 in Dataset-4. According to the SVM model, the most important features for Datasets 1–4 were conformation, chest width, meat yellowness, and C20:5 n3, respectively. In conclusion, this study provides a robust proof-of-concept for the high-accuracy classification of the investigated breeds using ML methods. The results suggest that LDA and SVM models based on carcass traits offer significant potential as objective tools for breed characterisation.