Semantic-Enhanced Graph Convolutional Neural Networks for Multi-Scale Urban Functional-Feature Identification Based on Human Mobility

Yuting Chen,Pengjun Zhao,Yi Lin,Yushi Sun,Rui Chen,Ling Yu,Yu Liu

DOI: https://doi.org/10.3390/ijgi13010027

IF: 3.4

2024-01-12

ISPRS International Journal of Geo-Information

Abstract:Precise identification of spatial unit functional features in the city is a pre-condition for urban planning and policy-making. However, inferring unknown attributes of urban spatial units from data mining of spatial interaction remains a challenge in geographic information science. Although neural-network approaches have been widely applied to this field, urban dynamics, spatial semantics, and their relationship with urban functional features have not been deeply discussed. To this end, we proposed semantic-enhanced graph convolutional neural networks (GCNNs) to facilitate the multi-scale embedding of urban spatial units, based on which the identification of urban land use is achieved by leveraging the characteristics of human mobility extracted from the largest mobile phone datasets to date. Given the heterogeneity of multi-modal spatial data, we introduced the combination of a systematic data-alignment method and a generative feature-fusion method for the robust construction of heterogeneous graphs, providing an adaptive solution to improve GCNNs' performance in node-classification tasks. Our work explicitly examined the scale effect on GCNN backbones, for the first time. The results prove that large-scale tasks are more sensitive to the directionality of spatial interaction, and small-scale tasks are more sensitive to the adjacency of spatial interaction. Quantitative experiments conducted in Shenzhen demonstrate the superior performance of our proposed framework compared to state-of-the-art methods. The best accuracy is achieved by the inductive GraphSAGE model at the scale of 250 m, exceeding the baseline by 25.4%. Furthermore, we innovatively explained the role of spatial-interaction factors in the identification of urban land use through the deep learning method.

geography, physical,remote sensing,computer science, information systems

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the accurate identification of the functional characteristics of urban spatial units, which is a prerequisite for urban planning and policy - making. However, inferring the unknown attributes of urban spatial units from spatial interaction data remains a challenge in Geographic Information Science. Although neural network methods have been widely used in this field, urban dynamics, spatial semantics and their relationships with urban functional characteristics have not been deeply explored.

To address these challenges, the authors propose Semantic - Enhanced Graph Convolutional Neural Networks (SE - GCNNs), aiming to achieve multi - scale embedding of urban spatial units through features extracted from human mobility data and, based on this, to achieve the identification of urban land use. Specifically, the paper addresses the following key issues:

1. **Heterogeneity of multi - modal spatial data**: A systematic data alignment method and a generative feature fusion method are introduced to robustly construct a heterogeneous graph, provide adaptive solutions, and improve the performance of GCNN in node classification tasks.
2. **Spatial scale effect**: For the first time, the influence of the spatial scale effect on the GCNN backbone network is explicitly examined. It is found that large - scale tasks are more sensitive to the directionality of spatial interactions, while small - scale tasks are more sensitive to the proximity of spatial interactions.
3. **Interpretability**: The role of spatial interaction factors in urban land use identification is innovatively explained through deep learning methods.

The paper verifies through quantitative experiments that the proposed framework is superior in performance to existing methods. In particular, at the 250 - meter scale, using the inductive GraphSAGE model, the accuracy is 25.4% higher than that of the baseline method. These contributions provide new perspectives and methods for the identification of urban functional characteristics.