@article{article_1797180, title={Predicting Honeybee Colony Health Using Weather and Apiary Data with Machine Learning}, journal={Türk Mühendislik Araştırma ve Eğitimi Dergisi}, volume={4}, pages={140–160}, year={2025}, author={Yılmaz, Ümit}, keywords={Honey bee health, healthy colony checklist, machine learning, ensemble methods}, abstract={Honey bee colonies are essential for global food security but continue to suffer heavy losses from interacting biological and environmental stressors. Predicting colony health is therefore a priority for sustainable apicultural management. This study examines the feasibility of forecasting honey bee colony health using weather and seasonal variables together with field assessments from the Healthy Colony Checklist. A dataset of 1,277 inspections from apiaries in North Carolina and Utah, integrated with meteorological records from nearby stations, was analyzed. Engineered features included vapor pressure deficit, temperature–humidity interactions, wind energy estimates, and seasonal encodings. The prediction was structured as a binary classification task (healthy vs. unhealthy). Several machine learning models were tested, emphasizing tree-based ensembles such as Random Forest, Extra Trees, Light Gradient Boosting Machine, Categorical Boosting, Gradient Boosting, Histogram-based Gradient Boosting, and Extreme Gradient Boosting. Ensemble strategies, including optimized soft voting and stacking, were also applied. Results showed accuracies of 75–76% with ROC AUC values near 0.80. Precision exceeded 0.70, while recall remained modest (~0.55). Seasonality was the dominant predictor, with weather indicators providing complementary value. Findings confirm the usefulness of agrometeorological data for decision-support in apiculture but also highlight the limits of weather-only models. Incorporating hive-level biological and management factors with advanced learning methods is recommended.}, number={2}, publisher={Türk Eğitim-Sen}