Relation-aware stock knowledge graph for interpretable urban renewal prioritization and scoring

We propose that urban renewal under a “stock-first” paradigm requires transparent and reproducible approaches to determine which existing blocks should be prioritized for optimization. This study introduces a relation-aware Stock Knowledge Graph (Stock-KG) to support interpretable composite scoring and the prioritization of urban renewal targets within the Digital Earth framework. The Stock-KG integrates 19 multisource urban datasets—including buildings, land-use zones, facilities, and more—into a typed multigraph centered on the PlanningBlock, connected by four spatial–semantic relations (WITHIN, OVERLAPS, ADJACENT, and EVALUATES). Five standardized components—intensity, land-use mix, facility accessibility, market activity, and adjacency synergy—are quantified and fused through entropy-based weighting, yielding transparent composite scores at the block level. Correlation diagnostics reveal that intensity, mix, and adjacency dominate renewal potential, while weak inter-component dependence confirms low redundancy. The resulting high-score blocks form coherent spatial clusters along adjacency backbones, enabling auditable identification of top-priority renewal areas. A backtracing mechanism further links each block’s score to its Top-K contributing objects, ensuring decision-level accountability. This relation-aware and interpretable scoring framework preserves the full evidence chain—from data to decisions—within confidential yet reproducible Digital Earth workflows. It provides a scalable methodological foundation for data-driven urban renewal and sustainable spatial governance.