Abstract:Spatial networks are useful for modeling geographic phenomena where spatial interaction plays an important role. To analyze the spatial networks and their internal structures, graph-based methods such as community detection have been widely used. Community detection aims to extract strongly connected components from the network and reveal the hidden relationships between nodes, but they usually do not involve the attribute information. To consider edge-based interactions and node attributes together, this study proposed a family of GeoAI-enhanced unsupervised community detection methods called region2vec based on Graph Attention Networks (GAT) and Graph Convolutional Networks (GCN). The region2vec methods generate node neural embeddings based on attribute similarity, geographic adjacency and spatial interactions, and then extract network communities based on node embeddings using agglomerative clustering. The proposed GeoAI-based methods are compared with multiple baselines and perform the best when one wants to maximize node attribute similarity and spatial interaction intensity simultaneously within the spatial network communities. It is further applied in the shortage area delineation problem in public health and demonstrates its promise in regionalization problems.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: How to simultaneously combine node attributes and edge connections (including spatial interactions and geographical adjacencies) in a spatial network to identify regions with similar attributes, strong spatial interactions, and geographical adjacency. Traditional community detection methods mainly rely on topological structures while ignoring node attribute information and have high computational costs when dealing with large - scale complex networks. In addition, although existing deep - learning methods can generate low - dimensional representations of nodes, their application in unsupervised learning tasks is limited, especially in the community detection tasks of spatial networks. To solve these problems, the authors propose an enhanced unsupervised community detection method based on GeoAI - region2vec. This method uses the Graph Attention Network (GAT) and the Graph Convolutional Network (GCN) to learn node embeddings and extracts communities through agglomerative clustering. Specifically, the main contributions of this study include: 1. **Combining Node Attributes and Edge Connections**: A community detection algorithm that can consider node attributes and edge connections (spatial interactions and geographical adjacencies) simultaneously is proposed, which overcomes the limitations of traditional methods. 2. **Designing a Loss Function for Community Detection**: To better adapt to unsupervised learning tasks, a new loss function is designed. This loss function includes spatial interaction flow constraints and spatial distance constraints, making nodes with spatial interactions closer in the embedding space, while nodes without spatial interactions are farther away. 3. **Application to Practical Problems**: The proposed region2vec method is applied to the problem of shortage area division in the field of public health, demonstrating its potential in regionalization problems. ### Formula Summary - **Forward Propagation Formula of the GCN Model**: \[ Z^{(1)}=\text{ReLU}(\tilde{D}^{-\frac{1}{2}}\tilde{A}\tilde{D}^{-\frac{1}{2}}XW^{(0)}) \] \[ Z^{(2)}=\tilde{D}^{-\frac{1}{2}}\tilde{A}\tilde{D}^{-\frac{1}{2}}Z^{(1)}W^{(1)} \] where \(\tilde{A} = A + I\) is the adjacency matrix after adding self - loops, and \(\tilde{D}\) is the corresponding degree matrix. - **Self - Attention Mechanism of the GAT Model**: \[ e_{ij}=a(W\mathbf{z}_i,W\mathbf{z}_j) \] \[ \alpha_{ij}=\text{softmax}_j(e_{ij})=\frac{\exp(e_{ij})}{\sum_{k\in N_i}\exp(e_{ik})} \] \[ \mathbf{z}'_i=\sigma\left(\sum_{j\in N_i}s'_{ij}\alpha_{ij}W\mathbf{z}_j\right) \] where \(s'_{ij}\) is the normalized traffic intensity coefficient. - **Loss Function**: \[ L_{\text{hops}}=\sum I(\text{hop}_{ij}>\epsilon)\frac{d_{ij}}{\log(\text{hop}_{ij})} \] \[ L = \frac{1}{N_{\text{pos}}}\sum_{p = 1}^{N_{\text{pos}}}\log(s_p)d_{\text{pos},p}-\frac{1}{N_{\text{neg}}}\sum_{q = 1}^{N_{\text{neg}}}d_{\text{neg},q}+L_{\text{hops}} \] Through these improvements, the region2vec method can perform community detection more effectively in spatial networks, especially when it is necessary to consider comprehensively.

GeoAI-Enhanced Community Detection on Spatial Networks with Graph Deep Learning

Region2Vec: Community Detection on Spatial Networks Using Graph Embedding with Node Attributes and Spatial Interactions

A Graph Convolutional Network-Based Approach for Community Detection in Attributed Networks

Detecting User Community in Sparse Domain via Cross-Graph Pairwise Learning

Graph Convolutional Networks Meet Markov Random Fields: Semi-Supervised Community Detection in Attribute Networks

Unsupervised Learning for Community Detection in Attributed Networks Based on Graph Convolutional Network

Sensing Overlapping Geospatial Communities from Human Movements Using Graph Affiliation Generation Models.

Attribute Diversity Aware Community Detection on Attributed Graphs Using Three-view Graph Attention Neural Networks

CommunityGAN: Community Detection with Generative Adversarial Nets

CommDGI

VGAER: Graph Neural Network Reconstruction based Community Detection

Query driven-graph neural networks for community search

Community Attention Network for Semi-supervised Node Classification.

Weakly-supervised learning for community detection based on graph convolution in attributed networks

Spatial Regression Graph Convolutional Neural Networks: A Deep Learning Paradigm for Spatial Multivariate Distributions.

Community detection in attributed social networks using deep learning

Community Detection Fusing Graph Attention Network

Community detection based on unsupervised attributed network embedding

Hierarchical attention network for attributed community detection of joint representation

Gumbel-SoftMax based graph convolution network approach for community detection

Unsupervised Graph Attention Autoencoder for Attributed Networks using K-means Loss