Unveiling optimal SDG pathways: an innovative automated recommendation approach integrating graph pruning, intent graph, and attention mechanism

The recommendation of Sustainable Development Pathways (SDPs) is crucial for achieving the United Nations Sustainable Development Goals (SDGs) at regional level. However, traditional recommendation algorithms struggle with two key challenges: spatial heterogeneity and sparse historical interaction records between regions and SDPs. To address these issues, we introduce the Regional Graph-Based Explainable Recommendation (RGB-ER) method. RGB-ER leverages a pruned Regional Graph (RG) to capture regional spatial heterogeneity, incorporating environmental, economic, and social factors into the recommendations. In addition, an Intent Graph models regional preferences across various attributes, bridging historical interactions with the RG and mitigating data sparsity. This dual approach significantly improves recommendation accuracy and interpretability. Extensive experiments show that RGB-ER outperforms state-of-the-art graph-based models, with a maximum improvement of 9.61% in Top-3 recommendation accuracy. A case study in Fujian Province–a region characterized by its mountainous terrain, complex socio-economic landscape, and significant sustainability challenges–illustrates RGB-ER’s practical applicability, aligning well with local government strategies for sustainable development. Furthermore, we assess SDPs at the county level across China, highlighting the method’s potential for guiding region-specific sustainable development planning. In conclusion, RGB-ER provides a robust, explainable framework for data-driven decision-making in sustainable development.