Architectural features of the implementation of a decision support system in grain crops yield management

The article presents architectural solutions for implementing a decision support system for grain crop yield management. The main objective is to integrate UAV data with historical and current GIS data to ensure adaptive yield prediction in real time. The proposed microservice architecture consists of three functional layers: data ingestion, processing and analytics, modeling and decisions. A key feature is the use of asynchronous communication through Apache Kafka message broker, which ensures loose coupling of components and high throughput. A service for monitoring the effectiveness of agrotechnical recommendations has been developed, which separates forecasting logic from business rule application logic. The system includes mathematical models for checking soil compaction conditions and assessing the feasibility of additional fertilization. Implementation of asynchronous approach ensures fault tolerance, scalability and independent service updates. The technology stack includes Python, scikit-learn, PyTorch, Django, Kubernetes, PostgreSQL/PostGIS. The result is decision support for management decisions on increasing grain crop yields in precision agriculture systems.