Prediction of daily egg-laying in japanese quails using machine learning

Accurate monitoring of individual productivity in quail farming is essential for maximizing economic returns and shaping effective, sustainable breeding programs. However, daily variation in egg-laying behavior across individuals presents significant challenges to consistent productivity prediction. While conventional methods often struggle to account for these irregular and long-term behavioral trends, machine learning techniques offer a promising alternative through their capacity for individualized, data-driven modeling. In this study, we propose a machine learning-based approach to forecast daily egg-laying outcomes by identifying production patterns from a longitudinal dataset comprising 193 days of continuous observations on 371 quails. The predictive feature set includes rolling averages, cumulative production metrics, and prior-day laying status engineered from each bird's historical production data. Three supervised classification models, Logistic Regression, Random Forest, and Extreme Gradient Boosting (XGBoost) were employed to predict egg-laying on a daily basis. Performance was assessed using accuracy, precision, recall, and F1-score. Among the models, XGBoost outperformed the others, achieving an F1-score of 91.2% and a recall of 97% on the test set. Feature importance analysis identified the 7-day rolling mean as the most influential predictor. These findings underscore the value of machine learning approaches in modeling individual-level laying patterns and demonstrate their potential application in selection decisions, flock-level management, and automated performance monitoring systems in poultry production.