Bringing realism: Enhancing high-dimensional data for active behavior analysis in older adults

Understanding active travel behavior among older adults in suburban communities is often hindered by restricted data access and imbalanced sample distributions at the community level. These challenges limit the ability to train robust nonlinear models that capture the complex relationships between t...

Full description

Saved in:
Bibliographic Details
Published in:Journal of transport geography Vol. 129; p. 104433
Main Authors: Guo, Cong, Jiang, Yaoqin, Yang, Yitong, Yuan, Zhilu, Guo, Renzhong
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.12.2025
Subjects:
Data augmentation
Generative adversarial networks
Light gradient boosting machine
Older adults travel behavior
Variational autoencoder
Variational autoencoder
Data augmentation
Generative adversarial networks
Older adults travel behavior
Light gradient boosting machine
ISSN:0966-6923
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Understanding active travel behavior among older adults in suburban communities is often hindered by restricted data access and imbalanced sample distributions at the community level. These challenges limit the ability to train robust nonlinear models that capture the complex relationships between travel behavior and built environment characteristics. To address these issues, a data generation algorithm is proposed based on 7631 data points from a case study of the household travel survey in Xiamen, China, comparing the Variational Autoencoder (VAE) and Generative Adversarial Networks (GANs) to overcome data collection constraints and enhance predictive model performance and generalization accuracy. Applying a gradient boosting algorithm to the meticulously generated dataset improved data set accuracy by 20 % for the VAE model and 13 % for the GAN model, indicating that VAE surpasses other data augmentation methods and demonstrates superior performance. This effective synthetic framework helps researchers and practitioners in the field of active travel by enabling the creation and use of synthetic mobility data, which can be leveraged to address real-world challenges associated with decision-making in active travel behaviors.
ISSN:0966-6923
DOI:10.1016/j.jtrangeo.2025.104433