Abstract

Background:
Creole goat husbandry for milk and meat improves food security in rural areas in Perú. Body weight (BW) is a key trait for selecting breeding stock, and it is estimated to be using algorithms. Likewise, BW is common in livestock farming.

Aim:
This study aimed to compare BW prediction models using a data mining algorithm in Creole goats, considering their biometric measurements.

Methods:
Data from 1,075 females aged between 1 and 4 years were used. Measurements of chest width, thoracic perimeter, wither height, sacrum height, rump width and length, body length, cannon bone perimeter, age, and region of the herd were recorded. The regression trees (classification and regression tree), support vector regression (SVR), and random forest regression (RFR) algorithms were used.

Results:
The SVR was better at predicting BWs in Creole goat herds. Similarly, the results were stable during training (R2 = 0.765) and testing (R2 = 0.707). However, it should be noted that RFR performed better with training data (R2 = 0.942).

Conclusion:
The proposed predictive models have demonstrated significant potential for accurately predicting BW based on biometric data. Finally, it contributes to better selection, feeding, and sanitary management of Creole goats.

Key words: Algorithms; Creole; Machine learning; Predictive models; Morphometrics goats.