Prediction of Tail Biting Events in Finisher Pigs from Automatically Recorded Sensor Data.

Area of Science:

• Animal Science
• Agricultural Engineering
• Machine Learning

Background:

• Tail biting in pigs is a significant animal welfare concern, leading to potential injury and economic losses.
• Early detection and intervention are crucial for preventing tail damage and improving pig welfare.
• Current methods for identifying tail biting events lack predictive capabilities for timely farmer intervention.

Purpose of the Study:

• To develop and validate a predictive algorithm for tail biting events in pigs using readily available sensor data.
• To assess the feasibility of using water usage and pen temperature data for early tail biting detection.
• To provide farmers with a tool for proactive management of tail biting in swine production.

Main Methods:

• Collected sensor data included water flow, water activation frequency, and pen temperature (solid and slatted floor).
• Employed dynamic linear models for data source modeling, artificial neural networks for optimization and training, and a Bayesian ensemble strategy for combining predictions.
• Validated the final prediction algorithm on a separate group of finisher pigs under real-life farming conditions.

Main Results:

• The developed prediction algorithm achieved an Area Under the Curve (AUC) greater than 0.80, indicating successful prediction of tail biting events.
• The algorithm demonstrated the potential to predict tail biting using existing sensor data.
• Approximately 30% of non-event days resulted in false alarms, highlighting a need for more specific predictors.

Conclusions:

• It is feasible to predict tail biting events in pigs using sensor data on water usage and pen temperature.
• The current algorithm shows promise for early warning but requires refinement to reduce false alarms.
• Farmers can utilize the generated alarms to identify pens requiring increased observation and potential intervention.