Spatial regression graph convolutional neural networks: A deep learning paradigm for spatial multivariate distributions

Keywords: Spatial regression; Graph convolutional neural networks; Deep learning; GeoAI; Social sensing
Session Type: Virtual Paper Abstract
Day: Tuesday
Session Start / End Time: 3/1/2022 03:40 PM (Eastern Time (US & Canada)) - 3/1/2022 05:00 PM (Eastern Time (US & Canada))
Room: Virtual 21

Abstract

Recently, graph-based deep learning has drawn great attention in the inter-discipline of computer science and geographical analysis, especially in modeling irregular spatial structures.
In this article, we set forth that spatial regression can be performed in the manner of graph convolutional neural networks (GCNNs), so as to overcome limitations such as the linear assumption and the weight matrix constraints in spatial econometrics.
We develop spatial regression graph convolutional neural networks (SRGCNNs) as a deep learning paradigm to cope with a wide range of geographical tasks where the spatial multivariate data needs modeling and prediction. The feasibility of SRGCNNs lies in the feature propagation mechanisms, the spatial locality nature, and a semi-supervised training strategy.
Experiments in the Beijing urban area are designed to test SRGCNNs in capturing the spatial relations between social media check-ins and the layout of urban points of interest (POIs). The results indicate that a well-trained SRGCNN model is capable of learning from samples and performing reasonable predictions for unobserved locations. Compared with benchmark spatial regression models, it is inspiring to find that SRGCNN-based models are less sensitive to the sampling ratio and could achieve better predictions when the observed data is insufficient.
This study offers to bridge the methodological gap between graph deep learning and spatial regression analytics. The proposed idea serves as an example to illustrate how spatial analytics can be combined with state-of-the-art deep learning models, and to enlighten future research at the front of geospatial artificial intelligence (GeoAI).