1. Optimizing Frost Prediction with a Multi-Window CNN–XGBoost Soft-Voting Ensemble

Recent global climate change has increased the frequency of late-spring frost events, causing more severe and widespread damage to orchard growers. Frost formation occurs due to rapid temperature drops over short periods combined with overnight air stagnation; thus, effective prediction requires analyzing patterns across multiple time scales. We introduce a hybrid frost-forecasting framework that combines a multi-window 1-D convolutional neural network (CNN), utilizing 6-, 12-, and 24-hour... D. Kim