Predicting body weight of male Kuroiler chickens from linear body measurements using MARS and CART data-mining algorithms.

Body weight is an essential trait in chickens, especially in markets where chickens are priced based on body weight and where medicine dosages depend on the animal's weight. However, not all farmers can afford scales to measure body weight, and they sometimes lack technical know-how during breakdowns, which is a challenge for small-scale farmers. Lately, data-mining algorithms have been used to help predict live body weight in livestock as they perform better than traditional prediction methods like linear regression. This study, therefore, aimed to develop models to predict the live body weight of the Kuroiler chicken breed from linear body measurements using classification and regression tree (CART) and multivariate adaptive regression spline (MARS) data-mining algorithms and to assess which of the two data-mining algorithms has a superior predictive performance. Linear body measurements were taken using a tailor's tape, and the body weight was taken using an electronic scale from 100 male Kuroiler chickens aged 23 weeks. The linear body measurements taken were corpus length (CL), chest circumference (CC), thigh length (TL), thigh circumference (TC), shank circumference (SC), shank length (SL), keel length (KL), and body length (BL). Results showed that the Kuroiler chickens used had an average live body weight of 2.01 kg, which correlated positively with all measured linear body measurements. Both MARS and CART developed models that included chest circumference, shank length, thigh circumference, and keel length. Furthermore, the MARS and CART models isolated the keel length and chest circumference as the most crucial traits in predicting the body weight of male Kuroiler chickens. The predictive performance showed that CART was the best model, with the highest r , R 2 , and adjusted R 2 . These results suggest that the CART data-mining algorithm might help to determine the breeding standards of the Kuroiler chicken breed for a breeding program. The findings of this study may be helpful to breeders, producers, and marketers of Kuroiler chickens.