Federated Logistics Operations Dataset (FLOD)

Abstract

The Federated Logistics Operations Dataset (FLOD) is a large-scale real-world dataset designed to support research on distributed logistics optimization, predictive modeling, and industrial Internet of Things (IIoT) analytics. The dataset consists of 253,020 operational records collected from geographically distributed logistics service providers operating across multiple urban and industrial regions. Each record represents a single logistics operation event, capturing a comprehensive snapshot of routing behavior, vehicle characteristics, energy usage, environmental conditions, and operational context. The data were aggregated from multiple decentralized logistics management systems, warehouse monitoring platforms, fleet telemetry sources, and environmental sensing infrastructures deployed at logistics hubs and transportation corridors. To reflect realistic industrial settings, the dataset preserves inherent heterogeneity and non-IID characteristics arising from regional variations, operational policies, fleet composition, and workload intensity across participating entities. No raw identifiers or sensitive business information are included, ensuring compatibility with privacy-preserving and federated learning research. FLOD contains 128 structured features spanning operational metrics (e.g., distance traveled, payload, idle time, stop frequency), fleet and energy indicators (e.g., vehicle type, fuel or energy consumption, battery health), routing and congestion factors, and environmental conditions such as weather severity and carbon intensity. Temporal context is provided through event indices, operational hours, and weekly cycles without exposing precise timestamps. The dataset further includes two continuous regression targets: the Sustainable Logistics Efficiency Index (SLEI), which quantifies overall operational efficiency, and the Carbon-Adjusted Delivery Cost (CADC), which reflects cost behavior adjusted for energy usage and carbon impact. The dataset is suitable for regression analysis, federated and decentralized learning, robustness evaluation under operational disturbances, and benchmarking of advanced machine learning models for intelligent logistics and IIoT systems.