Body Weight Prediction from Linear Measurements of Icelandic Foals: A Machine Learning Approach.

Area of Science:

• Equine Science
• Animal Biometrics
• Machine Learning Applications

Background:

• Accurate body weight measurement in horses is vital for optimal feeding, care, and health monitoring by breeders and veterinarians.
• Traditional methods like weighbridges are not always feasible, necessitating alternative estimation techniques.
• Biometric measurements offer a potential solution for non-invasive body weight estimation in horses.

Purpose of the Study:

• To develop and validate machine learning models for estimating body weight in Icelandic foals.
• To identify key biometric features for accurate weight prediction in young horses.
• To provide practical tools for estimating foal body weight without specialized equipment.

Main Methods:

• Utilized a dataset of 312 biometric measurements from 24 Icelandic foals (birth to 404 days).
• Developed and compared various machine learning models, including polynomial regression.
• Validated model performance using cross-validation and a holdout dataset.

Main Results:

• A polynomial model incorporating heart girth, body circumference, and cannon bone circumference achieved a mean percentage error of 4.1% (cross-validation) and 3.8% (holdout).
• A simpler model using the square of heart girth multiplied by body circumference demonstrated a mean percentage error of up to 5%.

Conclusions:

• Machine learning models provide a reliable and accurate method for estimating Icelandic foal body weight.
• Biometric-based formulas offer a practical alternative to weighbridges for routine monitoring.
• These findings support the utility of machine learning in developing predictive models for equine management.