Article Preview
TopIntroduction
In today's fast-changing global marketplace, supply chain collaboration and cooperation have become key to achieving efficiency and responsiveness (Qin & Guo, 2021). Supply chain collaboration and cooperation refers to the coordination and cooperation between different organizations in a supply chain to achieve a common goal (Hanga & Kovalchuk, 2019). These practices involve the coordination among organizations to optimize supply chain performance, which enhances efficiency, reduces costs, and improves customer satisfaction (Zhang et al., 2023). Driven by globalization and technological advances, supply chain collaboration and cooperation have become particularly important (Wankmüller &Reiner, 2019). However, achieving effective supply chain collaboration and cooperation faces many challenges, such as the security of information sharing, the conflict of goals among different organizations (LI Yueze, 2023), and the difficulty of data processing and analysis in complex supply chain structures.
In recent years, deep learning technology has emerged as a key solution to supply chain collaboration issues (Yuxiang et al., 2021). The ability of deep learning lies in that it could effectively extract information from complex data, deepening our understanding of supply chain dynamics, optimizing decision-making processes, and predicting trends (Liu et al., 2022). Deep learning models such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and graph neural networks (GNNs) have been used in a variety of areas such as demand forecasting (Nunes et al., 2020), inventory management, and supply chain risk assessment, demonstrating strong computational and analytical forecasting capabilities.
Based on the shortcomings of the above work, we introduce the Bidirectional Encoder Representation Transformer (BERT)-graph attention network (GAT)-reinforcement learning (RL) model to address supply chain collaboration challenges. The BERT-GAT-RL model aims to enhance data processing and decision-making in supply chains. In this model, the BERT part is responsible for processing and understanding supply chain textual data, such as communication records of suppliers and customers; the GAT part is used to analyze complex relationships and influences in the supply chain network; and the RL part explores optimal supply chain decision-making strategies in a simulation environment.
The innovation of our paper is the integration of the BERT, GAT, and RL models for a comprehensive approach to challenges in supply chains and the ability to deal with multiple issues in supply chain collaboration and cooperation in a more holistic way. This combination leads to more accurate and efficient supply chain decisions and promotes better collaboration among stakeholders. BERT is used for processing textual data within the supply chain, such as communication records. Its bidirectional nature allows it to capture contextual information effectively, improving the efficiency of using textual data in supply chain management. GAT is employed to analyze complex relationships and influences within the supply chain network. By leveraging attention mechanisms, GAT can focus on relevant nodes in the network, aiding in the structuring of complex supply chain relationships. RL is utilized to explore optimal supply chain decision-making strategies in a simulation environment. RL's ability to learn from interactions with the environment enables it to adapt and optimize decisions in dynamic supply chain scenarios.