Predicting veal-calf trading events in France

Area of Science:

• Veterinary Epidemiology
• Animal Health Economics
• Statistical Modeling

Background:

• Global trade is a major driver of infectious disease spread.
• Livestock trade accelerates disease circulation and long-distance dissemination.
• Predicting animal movements is crucial for epidemic risk assessment.

Purpose of the Study:

• Develop a statistical framework to predict livestock trading events.
• Utilize routinely collected data for predictive modeling.
• Assess the impact of data acquisition delays and prediction lag times.

Main Methods:

• Employed a random forest binary classification model.
• Focused on veal calf trade data in France (2011-2019).
• Analyzed trade events using municipality as the trading unit and a 28-day period.

Main Results:

• Sales events were predicted more accurately than purchase events.
• Model performance decreased with delayed data acquisition; prediction lag had minimal impact.
• The veal calf trading network appears to be sales-driven.

Conclusions:

• The developed random forest model effectively predicts veal calf trading events.
• Optimal performance was achieved using municipality-level data and a 28-day trading period.
• Predicted trade networks can enhance pathogen spread simulations in livestock.