A Multi-Source Data Fusion Framework for Resilient Demand Forecasting in Smart Logistics

Abstract

Smart logistics systems increasingly depend on precise demand forecasting to streamline global supply chain operations. Current industry data reveals that logistics bottlenecks impact more than 30% of global shipments, while inaccurate forecasting contributes to a 25% surge in operational expenditures. Conventional manual estimation techniques are often inadequate, as they struggle to incorporate volatile real-world variables such as fluctuating traffic patterns, IoT telemetry, and environmental shifts. This research introduces a robust data fusion framework that synthesizes IoT, traffic, and meteorological datasets to refine demand forecasting and logistics delay projections via a hybrid Classification and Regression Tree (CART) approach. The methodology initiates with a rigorous preprocessing phase where heterogeneous data streams are cleaned, normalized, and temporally synchronized. While baseline models such as K-Nearest Neighbor (KNN) and Categorical Boosting (CatBoost) offer foundational insights, they frequently overlook the intricate interdependencies inherent in multi-source data. To address these limitations, this study proposes the Tao Tree model. This architecture utilizes a Tao Tree module for hierarchical feature selection and adaptive weighting, paired with a CART module to deliver high-precision regression for demand levels and categorical delay assessments. The integrated system is deployed via a Flask-based web application, facilitating real-time data ingestion and predictive visualization. Experimental results indicate that this framework substantially elevates forecasting accuracy and operational efficiency, offering a scalable, data-driven solution for proactive supply chain management.