Journal
INTERNATIONAL JOURNAL OF GEOGRAPHICAL INFORMATION SCIENCE
Volume -, Issue -, Pages -Publisher
TAYLOR & FRANCIS LTD
DOI: 10.1080/13658816.2023.2254382
Keywords
Geodemographic classification; clustering; deep learning; graph convolutional neural networks
Ask authors/readers for more resources
This study presents a systematic investigation on the use of graph neural networks for geodemographic classification. Using Greater London as a case study, the results show that the proposed Node Attributes-focused Graph AutoEncoder framework performs well in terms of class homogeneity and spatial clustering.
Geodemographic classifications are exceptional tools for geographic analysis, business and policy-making, providing an overview of the socio-demographic structure of a region by creating an unsupervised, bottom-up classification of its areas based on a large set of variables. Classic approaches can require time-consuming preprocessing of input variables and are frequently a-spatial processes. In this study, we present a groundbreaking, systematic investigation of the use of graph neural networks for spatial geodemographic classification. Using Greater London as a case study, we compare a range of graph autoencoder designs with the official London Output Area Classification and baseline classifications developed using spatial fuzzy c-means. The results show that our framework based on a Node Attributes-focused Graph AutoEncoder (NAGAE) can perform similarly to classic approaches on class homogeneity metrics while providing higher spatial clustering. We conclude by discussing the current limitations of the proposed framework and its potential to develop into a new paradigm for creating a range of geodemographic classifications, from simple, local ones to complex classifications able to incorporate a range of spatial relationships into the process.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available