Broiler chickens can benefit from machine learning: support vector machine analysis of observational epidemiological...

Area of Science:

• Veterinary epidemiology
• Poultry science
• Machine learning applications

Background:

• Hock burn is a significant welfare issue in commercial broiler production.
• Routinely collected farm management data can be leveraged for health analysis.
• Machine learning offers advanced analytical capabilities for epidemiological data.

Purpose of the Study:

• To apply support vector machine learning to identify features associated with hock burn.
• To assess the predictive accuracy of machine learning for high hock burn prevalence.
• To compare machine learning with traditional logistic regression for this application.

Main Methods:

• Utilized support vector machine learning algorithms.
• Analyzed routinely collected commercial broiler farm management data.
• Compared results with multivariable logistic regression.

Main Results:

• Developed a machine learning classifier to predict hock burn occurrence.
• Achieved an accuracy of 0.78 (Area Under the ROC Curve) on unseen data.
• Identified key features associated with hock burn prevalence.

Conclusions:

• Machine learning, specifically support vector machines, effectively identifies hock burn risk factors.
• This technique provides novel insights beyond traditional statistical methods.
• Integrating machine learning into poultry management systems can enhance broiler health and welfare globally.