Coupling human mobility and social relationships to predict individual socioeconomic status: A graph neural network approach

Understanding individual's socioeconomic status (SES) can provide supporting information for designing political and economic policies. Acquiring large‐scale economic survey data is time‐consuming and laborious. The widespread mobile phone data, which can reflect human mobility and social netwo...

Full description

Saved in:
Bibliographic Details
Published in:Transactions in GIS 2024-08, Vol.28 (5), p.1412-1438
Main Authors: Chen, Xiao, Pei, Tao, Song, Ci, Shu, Hua, Guo, Sihui, Wang, Xi, Liu, Yaxi, Chen, Jie
Format: Article
Language:English
Subjects:
Cell phones
Cellular telephones
Coupling
Data acquisition
Economic policy
Economics
Feature extraction
Geography
Graph neural networks
Information processing
Mobility
Modules
Neural networks
Social networks
Social organization
Socioeconomic factors
Socioeconomic status
Socioeconomics
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Understanding individual's socioeconomic status (SES) can provide supporting information for designing political and economic policies. Acquiring large‐scale economic survey data is time‐consuming and laborious. The widespread mobile phone data, which can reflect human mobility and social network characteristics, has become a low‐cost data source for researchers to infer SES. However, previous studies often oversimplify human mobility features and social network features extracted from mobile phone data into general statistical features, resulting in discounting some important temporal and relational information. Therefore, we propose a comprehensive framework for individual SES prediction that effectively utilizes a combination of human mobility and social relationships. In this framework, Word2Vec module extracts human mobility features from mobile phone positioning data, and graph neural network (GNN) module GraphSAGE captures social network characteristics constructed from call detail records. We evaluated the effectiveness of our proposed approach by training the model with real‐world data in Beijing. According to the experimental results, our proposed hybrid approach outperformed the other methods evidently, demonstrating that human mobility and social links are complementary in the characterization of SES. Coupling human mobility and social links can further deepen our understanding of cities' economic geography.
ISSN:1361-1682
1467-9671
DOI:10.1111/tgis.13189