Predicting the body weight of crossbred Holstein × Zebu dairy cows using multivariate adaptive regression splines...

Area of Science:

• Animal Science
• Agricultural Engineering
• Biometrics

Background:

• Accurate estimation of live body weight is crucial for dairy cattle management.
• Traditional methods of weighing can be impractical in tropical environments.
• Developing non-invasive prediction methods is essential for efficient livestock management.

Purpose of the Study:

• To estimate live body weight in Holstein × Zebu dairy cows using readily available body measurements.
• To evaluate the effectiveness of the multivariate adaptive regression splines (MARS) algorithm for body weight prediction.
• To provide a reliable tool for animal breeders and researchers in tropical regions.

Main Methods:

• Utilized body measurements including height, width, and girth from 156 Holstein × Zebu dairy cows.
• Employed the multivariate adaptive regression splines (MARS) algorithm for model construction.
• Applied various train-test data proportions (65:35, 70:30, 80:20) to validate the prediction model.

Main Results:

• The MARS algorithm demonstrated strong predictive capability, with an explanation rate of 0.836 for the 80:20 train-test set.
• The model achieved minimum Akaike information criterion values, indicating good model fit.
• The algorithm proved to be a reliable method for predicting body weight from body measurements.

Conclusions:

• The MARS algorithm offers a reliable and accurate approach for estimating dairy cow body weight.
• This predictive tool can significantly support animal breeders and researchers in optimizing feeding and selection programs.
• The study highlights the utility of MARS in agricultural applications for livestock management.