دانشگاه آزاد اسلامی واحد تهران جنوب

صفحه اصلی / سومین کنفرانس ملی فناوری های نوین در انرژی

Multi-Criteria Attention-Based Graph Neural Network: A Heterogeneous Representation Learning Framework for Materials Logistics System Optimization

نویسندگان :

Mohammad Shahbazi¹ Hamid Tohidi² Majid Nojavan³

1- School of Industrial Engineering, South Tehran Branch Islamic Azad University 2- School of Industrial Engineering, South Tehran Branch Islamic Azad University 3- School of Industrial Engineering, South Tehran Branch Islamic Azad University

کلمات کلیدی :

Machine learning،Deep learning،Representation learning،Heterogeneous systems،Logistics optimization،Materials logistics

چکیده :

need has emerged to model the intricate relationships within complex materials logistics systems in order to optimize routing, scheduling, and distribution in modern supply chains that manage diverse categories of materials. These systems often exhibit considerable variability in their facilities, transportation modes, and capacity constraints, creating what is referred to as ”heterogeneity,” which complicates the modeling process. To simplify calculations, some approaches assume homogeneous systems, thus neglecting critical variability in nodes (such as warehouses and distribution centers) and edges (such as transportation routes and capacity). However, overlooking this heterogeneity can result in marked reductions in accuracy. A representation learning method specifically tailored for heterogeneous materials logistics systems is proposed here. This method is designed to preserve multifaceted relationships among components and to improve model performance in real-world scenarios. Two novel extensions refine the underlying graph-based deep learning architecture by incorporating elements from deep learning, graph probability models, and machine learning. The approach is evaluated on a representative logistics network dataset using precision, F1 score, and recall as performance metrics. The experimental results indicate that this method outperforms existing solutions by providing higher precision, more accurate classification of system components, and better extraction of relationships within complex logistics networks.