A spatial–temporal dynamic attention-based Mamba model for multi-type passenger demand prediction in multimodal public transit systems

Authors

• Zhiqi Shao
• Haoning Xi
• David A. Hensher
• Ze Wang
• Xiaolin Gong
• Junbin Gao

Document Type

Journal Article

Publication Date

2025

Subject Area

place - australasia, place - urban, mode - bus, mode - ferry, mode - rail, ridership - demand, ridership - young people, ridership - old people, ridership - modelling, policy - equity, policy - sustainable

Keywords

Multimodal public transit systems, Multi-type passenger demand prediction, STDAtt-Mamba, Spatial–temporal dynamic fusion, Sparse attention, AI and deep learning

Abstract

Predicting passenger demand across multiple socio-demographic groups, such as adults, seniors, pensioners, and students, is essential for improving the operational efficiency, equity, inclusivity, and sustainability of multimodal public transit (PT) systems. Traditional demand prediction models, however, often fail to effectively capture the complex spatial–temporal variability inherent in heterogeneous socio-demographic groups. To bridge this gap, we propose a novel spatial–temporal dynamic attention-based state–space model, i.e., STDAtt-Mamba, tailored for multi-type passenger demand prediction in multimodal PT systems. The proposed STDAtt-Mamba model consists of three key components: an adaptive embedding layer that integrates station-level, passenger-type-specific, and temporal embeddings into a unified representation for efficient data processing; a spatial–temporal dynamic attention (STDAtt) module that employs sparse attention mechanisms to selectively capture crucial global spatial–temporal dynamics; and a spatial–temporal dynamic Mamba (STDMamba) module that extends the state–space modeling framework to fuse spatial and temporal dependencies dynamically. We prove that STDAtt-Mamba is a kind of spatial–temporal dual-path attention mechanism and theoretically validate the complementarity of STDMamba and STDAtt in capturing local and global dependencies, thereby improving the interpretability of the proposed STDAtt-Mamba. Extensive experiments are conducted on a large-scale multimodal PT dataset, including over 1.58 million passengers across nine distinct passenger groups (i.e., adults, seniors, pensioners, tertiary students, children, job seekers, school passengers, youth, and Gold Repat passengers) using travel modes such as bus, rail, and ferry, in Queensland, Australia, from January 2021 to January 2023. Experimental results demonstrate that the prediction performance of the proposed STDAtt-Mamba is superior to the19 baseline models with manageable computational costs, setting it as a state-of-the-art benchmark model for predicting the multi-type passenger demand in multimodal PT systems. This study offers an adaptive, scalable, robust, inclusive, and efficient predictive tool for transit authorities.

Rights

Permission to publish the abstract has been given by Elsevier, copyright remains with them.

Comments

Transportation Research Part E Home Page:

http://www.sciencedirect.com/science/journal/13665545

Recommended Citation

Shao, Z., Xi, H., Hensher, D. A., Wang, Z., Gong, X., & Gao, J. (2025). A spatial–temporal dynamic attention-based Mamba model for multi-type passenger demand prediction in multimodal public transit systems. Transportation Research Part E: Logistics and Transportation Review, 202, 104282.