Federated Learning-Enhanced Query Optimization for Federated Data Warehouses

Federated Data Warehouses (FDWs) integrate heterogeneous and distributed data sources, enabling unified query access without centralizing storage. However, query optimization in FDWs faces challenges due to data distribution, network latency, and varying source capabilities. This research proposes a novel Federated Learning-Enhanced Query Optimization (FLEQO) framework that leverages distributed machine learning to collaboratively learn optimal query execution strategies across multiple data sources while preserving data privacy. The framework employs model aggregation to capture execution cost patterns, adaptive join reordering, and source selection policies without transferring raw data. Experimental evaluation on synthetic benchmarks and real-world healthcare and financial datasets demonstrates significant improvements in query latency (up to 37% reduction) and execution cost estimation accuracy (up to 22% improvement) compared to traditional cost-based and heuristic methods. The results show that federated learning can enhance query optimization in FDWs by enabling adaptive, privacy-preserving, and cross-domain performance tuning, making it suitable for sensitive and large-scale enterprise environments.