Abstract
Amid growing global food security challenges and climate change, big data-driven crop prediction and breeding decisions are key to sustainable agriculture. By integrating multisource data, such as remote sensing, weather, soil, phenotypes, and genomics, with machine learning, this approach improves yield prediction accuracy and deepens understanding of genotype-by-environment (G×E) interactions. This paper summarizes recent advances in crop performance modeling and intelligent breeding, highlighting strategies for data integration, model generalization, environmental response modeling, and nonlinear phenotypic regulation. Emerging algorithms like deep learning, transfer learning, and federated learning enhance model adaptability, while cross-scale simulations enable dynamic agricultural regulation. The coupling of genomic selection with environmental data underpins intelligent breeding decisions, supporting precision agriculture and sustainable development.
