Kafkas Üniversitesi Veteriner Fakültesi Dergisi 2022 , Vol 28 , Issue 3
Phenotypic Characterization of Hair and Honamli Goats Using Classification Tree Algorithms and Multivariate Adaptive Regression Spline (MARS)
Yasin ALTAY1
1Eskisehir Osmangazi University, Faculty of Agriculture, Department of Animal Science, Biometry and Genetics Unit, TR-26160 Eskişehir - TÜRKİYE DOI : 10.9775/kvfd.2022.27163 Some morphological and physiological data are needed to scientifically describe animals and distinguish breeds from one another. Except for those who are not experts in the field, it is difficult to distinguish goat breeds from each other. Using data mining algorithms, this study aimed to develop a new phenotypic characterization for Honamli and Hair goats via some body measurement characteristics. In the study, some body characteristics of the Hair goat (65 animals) and the Honamli goat (83 animals) were used as independent variables. Th e dependent variable of the data mining algorithms, on the other hand, was defined as the binary response variable of Honamli and Hair breeds. Th e success of the CHAID, Exhaustive CHAID, CART, QUEST, and MARS algorithms in breed discrimination was determined at 87.80%, 85.80%, 87.80%, 77.00%, and 88.51%, respectively, while the area under the ROC curve was detected 0.880, 0.853, 0.868, 0.784, and 0.942, respectively, and Cohen"s Kappa coefficient (κ) 0.755, 0.711, 0.749, 0.549 and 0.739, respectively. As a result, the phenotype characterization of Honamli and Hair goats, whose morphological distinctions could not be made exactly, in MARS and CHAID algorithms, achieved with high success compared to other methods. Th e present study showed that Honamli and Hair goats may be distinguished by suitable statistical algorithms based on morphological data, which can be integrated with goat breeding studies to detect the origin of breeding animals. Keywords : CART, CHAID, Classification, Exhaustive CHAID, MARS, QUEST