Climate resilience analysis of public spaces via model-based artificial intelligence methods

Galiger, Gergő and Chien, Nguyen Duy and Hideg, Viktória and Tóth, Patrik and Kovács, Péter (2026) Climate resilience analysis of public spaces via model-based artificial intelligence methods. In: Re-Generation in Transport, Proceedings of the 11th TRA Conference. Lecture Notes in Mobility . Springer, Budapest. ISBN 978-3-032-37192-8 (In Press)

Text E2.pdf - Accepted Version Restricted to Registered users only Download (1MB)

Abstract

The efficient planning of climate-resilient infrastructure supposes an in-depth understanding of current and future mobility patterns, which are highly influenced by climatic factors in the case of public spaces. Recently, machine learning (ML) techniques, particularly neural networks (NN), have been proven efficient for analyzing and forecasting such complex traffic patterns. However, their large-scale adoption is limited by the underlying difficulties in pedestrian data collection and the lack of interpretability of such black-box approaches. In this study, we propose a data-efficient NN framework for analyzing and forecasting mobility patterns in public spaces. First, we employ a data collection method based on computer vision to measure the number of visitors on the Pest-side lower embankment of Budapest. Then, we extrapolate the obtained data for training a model-based NN to predict the number of visitors according to changes in temperature conditions. The proposed VPNgBTemp architecture combines variable projection networks (VPNet) with the nonlinear grey Bernoulli (NgB) model, explicitly capturing the inverted U-shaped trajectories of the visitor and temperature data. Finally, we show that VPNgBTemp is capable of modeling complex interactions of such weather conditions and mobility patterns, providing interpretable insights for databased climate resilience evaluation. Source code is available at: github.com/galigergergo/VPNgBTemp.

Actions (login required)

View Item