PageRank-Enhanced Deep Graph Network for Green Logistics Impact Assessment

Globalization and sustainability pressures demand accurate evaluation of green logistics supply chains. However, traditional methods are hampered by diverse data and complex relationships that limit their accuracy. To resolve these problems, this paper proposes a novel model called the deep graph network with PageRank-based feature selection combined with the meta-path-based heterogeneous graph convolutional network. The proposed approach integrates diverse indicators from supply chain management, operations, and technology. First, the PageRank algorithm selects key features, reduces data dimensionality, and retains essential information. Next, meta-path-based heterogeneous graph convolutional network explores and models complex relationships between these features using heterogeneous graph convolution and meta-paths. This process enables precise predictions of environmental metrics. Experimental results show that deep graph network with PageRank-based feature selection significantly improves prediction accuracy and robustness. Overall, this model provides an effective tool for evaluating and optimizing green supply chains.