Recursive prediction of broiler growth response to feed intake by using a time-variant parameter estimation method

Area of Science:

• Animal Science
• Agricultural Engineering
• Computational Biology

Background:

• Accurate prediction of broiler chicken growth is crucial for efficient farm management and resource allocation.
• Dynamic growth responses to feed intake necessitate adaptive modeling approaches.
• Existing static models may not fully capture the real-time, time-variant nature of broiler growth.

Purpose of the Study:

• To evaluate the efficacy of time-variant parameter estimation for real-time broiler growth modeling and prediction.
• To determine the optimal window size for recursive linear modeling of broiler growth.
• To compare the predictive accuracy of the recursive approach against static empirical growth models.

Main Methods:

• Utilized a recursive linear model with parameters estimated every 24 hours using a fixed time window of measurements.
• Analyzed 48 broiler chicken datasets to assess model performance.
• Compared prediction errors (Mean Relative Prediction Error - MRPE) with linear and nonlinear static growth models.

Main Results:

• The recursive linear modeling approach achieved minimum MRPE with a 5-day window size.
• Bird weight could be predicted 3 to 7 days ahead with an MRPE of 5% or less.
• The recursive model showed similar accuracy to nonlinear static models (1.4–2.3% MRPE vs. 1.1–2.8% MRPE) but was less accurate for longer prediction horizons (2–7 days).

Conclusions:

• Recursive modeling accurately predicts broiler growth in real time without prior system knowledge.
• The approach accounts for the time-variant, nonlinear growth process using limited data.
• This method is suitable for real-time integration into broiler production process management systems.