Interpreting core forms of urban morphology linked to urban functions with explainable graph neural network

2025-02-22

AI Summary

Urban functional identification faces challenges in modeling high-order morphological–functional relationships and ensuring model interpretability. Method: We propose an interpretable graph neural network (GNN) framework featuring a morphology–function attribution path decoupling mechanism. A morphology-topology encoder captures structural features—e.g., block density and road-network connectivity—while counterfactual perturbation analysis enables spatially explicit, semantically aligned attribution. A multi-city graph atlas supports cross-regional generalization. Contribution/Results: This work achieves, for the first time, unified structural interpretability and semantic understandability of GNNs for urban functional inference. Evaluated on six major cities, our method attains an F1-score of 0.89 for functional classification and >82% accuracy in attributing critical morphological elements. It further generates a verifiable, transferable morphology–function mapping rule base, advancing both transparency and practical applicability in urban analytics.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Problem

• Explainable graph neural network for urban morphology
• High-order relationship between urban form and function
• Core urban morphology representation (CoMo)

Innovation

• Explainable graph neural network
• Core urban morphology representation
• Deep learning framework CoMo