An Initial Study on the Use of Machine Learning and Radio Frequency Identification Data for Predicting Health...

Area of Science:

• Animal Science
• Agricultural Technology
• Data Science

Background:

• Free-range systems present unique health challenges for laying hens, impacting productivity and consumer confidence.
• Forecasting individual hen health can lead to significant economic benefits by reducing mortality and increasing egg production.
• Flock monitoring systems offer a viable solution for proactive health management in poultry.

Purpose of the Study:

• To investigate the use of Radio Frequency Identification (RFID) technologies and machine learning for predicting laying hen health status.
• To identify production system usage patterns and forecast health challenges throughout the production lifecycle.
• To analyze data for correlations and structures significant for predictive model performance on imbalanced datasets.

Main Methods:

• Development of a machine learning workflow incorporating data resampling to address dataset imbalance.
• Identification and refinement of important data features for predictive modeling.
• Utilizing Radio Frequency Identification (RFID) data to track individual hen behavior and patterns.

Main Results:

• The study achieved promising predictive performance for common laying hen diseases.
• An average of 28% of Spotty Liver Disease, 33% of roundworm, and 33% of tapeworm infections were correctly predicted.
• Monitoring hens in the early stages of egg production yielded similar predictive performance to later stages.

Conclusions:

• RFID and machine learning show potential for forecasting individual laying hen health in free-range systems.
• Early-stage monitoring provides valuable data for predictive health models.
• Incorporating additional data streams could further enhance the accuracy of flock health predictions.