Forecasting egg production in free-range laying hens using multi-farm data.

Area of Science:

• Agricultural Science
• Animal Science
• Data Science

Background:

• Accurate forecasting of laying rates in poultry, including sudden drops, is challenging despite technological progress.
• Commercial free-range hen farms require effective tools for anticipating production changes and optimizing management.

Purpose of the Study:

• To evaluate predictive modeling approaches for forecasting laying rates and detecting reduced production days.
• To assess the potential of integrating data from multiple commercial free-range farms for improved prediction.

Main Methods:

• Analysis of historical production and weather data from four commercial free-range farms (106 flocks, 35,346 flock-days).
• Development of three Random Forest models: single-farm, multi-farm, and combined.
• Performance evaluation using Area Under the Receiver Operating Characteristic Curve (AUC) for classification and Root Mean Squared Error (RMSE) for regression.

Main Results:

• The single-farm model achieved a median AUC of 0.86 and RMSE of 2.8.
• The multi-farm model showed higher AUC (0.89) but lower regression accuracy (RMSE = 4.98).
• The combined model improved regression (RMSE = 2.55) but had the lowest classification accuracy (AUC = 0.82).

Conclusions:

• Models developed on one farm can be effectively applied to others, indicating potential for cross-farm prediction.
• Integrating multi-farm data can support egg production forecasting in free-range systems.
• Combining datasets does not consistently improve model performance, but the approach aids farm management.