Estimation of gestating sows' welfare status based on machine learning methods and behavioral data

Area of Science:

• Animal Science
• Machine Learning Applications
• Livestock Management

Background:

• Individual welfare assessment in farms is crucial for livestock monitoring.
• Emerging technologies enable livestock behavior analysis using machine learning and sensors.

Purpose of the Study:

• To estimate components of gestating sows' welfare status using machine learning and behavioral data.
• To develop an automatic decision support system for sow welfare categorization.

Main Methods:

• Utilized a dataset combining individual and group behavioral measures (activity, social, feeding).
• Employed a clustering method to categorize welfare status for 69 sows over gestation periods.
• Developed a decision tree classifier using data from automatic feeders and video analysis.

Main Results:

• Identified three sow welfare clusters: 'scapegoat,' 'gentle,' and 'aggressive.'
• Found that environmental conditions and sow health influenced cluster proportions, not individual sow characteristics.
• Confirmed the significant impact of group behavior on individual sow welfare.
• Achieved over 72% accuracy in classifying sows into welfare categories.

Conclusions:

• Machine learning and behavioral analysis offer a promising approach for automated sow welfare assessment.
• The developed system can categorize gestating sow welfare based on individual and group behavior.
• This technology supports improved livestock animal monitoring and welfare management.