Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Civil and Environmental Engineering | en_US |
| dc.contributor.advisor | Ma, Wei (CEE) | en_US |
| dc.contributor.advisor | Chen, Anthony (CEE) | en_US |
| dc.creator | Li, Mingxi | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/13799 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Overcoming data and model limitations in practical traffic prediction for intelligent transportation systems | en_US |
| dcterms.abstract | Intelligent Transportation Systems (ITS) leverage advanced technologies to enhance the efficiency, safety, and sustainability of transportation networks. A critical component of ITS is spatiotemporal traffic prediction, which enables efficient management and optimization of traffic flow, path planning, and resource allocation. This thesis addresses practical challenges in spatiotemporal traffic prediction, focusing on both model and data limitations. | en_US |
| dcterms.abstract | While numerous computational models have been developed for general traffic prediction, most operate under idealized assumptions. In practice, real-world traffic prediction and its applications face significant non-idealities. To bridge this gap, we propose novel learning frameworks that integrate AI methodologies with domain knowledge to address critical data and model limitations. Overcoming data and model limitations is crucial for practical traffic prediction. Missing or insufficient data reduces accuracy, while opaque or inefficient models hinder real-world applications. Addressing these challenges improves prediction robustness and interpretability, enabling smarter traffic management. This research bridges theoretical advancements with practical applications, such as dynamic path planning, benefiting urban mobility. | en_US |
| dcterms.abstract | To address these challenges, this thesis investigates advanced methods that mitigate both model and data limitations, ultimately enhancing the accuracy and applicability of spatiotemporal traffic predictions. The thesis begins by examining the issues of data limitations, including sparse data and few archived data. First, the missing observations problem in traffic prediction is addressed through a novel deep learning model with a self-assisted imputation mechanism, which strengthens the robustness of predictions, even when confronted with inconsistent or incomplete datasets. Second, the thesis investigates the few-sample traffic prediction to determine the minimum data requirements for achieving accurate predictions. The deep learning model, equipped with relational inductive biases, effectively captures complex spatiotemporal patterns, making it particularly suited for environments with limited data. | en_US |
| dcterms.abstract | The model limitations are further explored with a focus on improving the interpretability and applicability of traffic prediction models. The interpretability of traffic prediction models is studied. A symbolic regression model is utilized to uncover implicit rules within traffic data variations, while simultaneously enhancing scalability and adaptability within ITS. Additionally, symbolic regression improves computational efficiency by deriving implicit equations that govern traffic dynamics. Furthermore, the integration of traffic prediction with path planning optimization is explored. Spatiotemporal predictions are leveraged to optimize real-time path planning, reducing travel times. This integration allows travelers to make informed decisions based on predicted traffic conditions within an end-to-end architecture, thereby enhancing travelers' mobility. | en_US |
| dcterms.abstract | These approaches are integrated into a unified, machine-learning framework for spatiotemporal traffic prediction under various challenging scenarios, including integration with path planning. By addressing the practical challenges of both model limitations and data limitations, the proposed framework enhances the interpretability, accuracy, and efficiency of traffic prediction. This, in turn, improves traffic management services and enhances travelers' mobility, ultimately making a significant contribution to the advancement of ITS. | en_US |
| dcterms.extent | xiv, 153 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2025 | en_US |
| dcterms.educationalLevel | Ph.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.LCSH | Intelligent transportation systems | en_US |
| dcterms.LCSH | Traffic estimation | en_US |
| dcterms.LCSH | Machine learning | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/13799