| Author: | Tang, Haoran |
| Title: | Temporal graph learning for dynamic recommendation |
| Advisors: | Li, Qing (COMP) |
| Degree: | Ph.D. |
| Year: | 2026 |
| Subject: | Recommender systems (Information filtering) Graph theory -- Data processing Neural networks (Computer science) Machine learning Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Computing |
| Pages: | xvi, 179 pages : color illustrations |
| Language: | English |
| Abstract: | Dynamic recommendation, where users continuously interact with items over time, has been widely deployed in real-world applications, such as social networks, streaming media, and online shopping platforms. Compared with the conventional static recommendation that follows the one-time recommendation-evaluation paradigm, dynamic recommendation involves real-time updates of user and item representations to refresh recommendation results and support multi-round recommendation process. Temporal graph learning, which constructs temporal interaction graphs using their sequential information, has demonstrated its superiority in modeling temporal representations and exploring dynamic preferences. Thus, this research focuses on fully investigating the power of temporal graph learning for dynamic recommendation. However, designing a comprehensive temporal update for user-item interaction graph still remains a challenge, since existing studies on temporal graph learning are mainly centered on some single insufficient perspectives for update. Moreover, they directly conduct representation update without scrutinizing whether the interactions would truly benefit temporal graph learning, which potentially degrades the recommendation performance. Thus, uncovering the effect of both newly-arrived and historical interactions poses another challenge to temporal update. In addition to the inherent challenges above, the unfairness issue among item exposures and user experiences tends to be amplified due to the evolving context in the dynamic environment. Unfortunately, most fairness-aware methods are designed for static recommendation. To address the mentioned challenges, this research first proposes a novel graph framework that comprehensively captures the dynamics of users and items from multiple perspectives, thus fully modeling the dynamic graph evolution for recommendation. Second, this research presents a novel temporal collaboration-aware graph network by investigating the collaborative effectiveness of the newly-arrived interactions. Third, this research constructs a novel self-correcting interaction network by dynamically filtering out the inappropriate historical interactions, to ensure the accuracy of temporal graph aggregation. In addition, this research further studies the problem of dual-side unfairness for the dynamic recommendation and presents an adaptive model-agnostic post-ranking method to mitigate the dynamic unfairness issue. Extensive experiments over real-world datasets demonstrate both the effectiveness and superiority of this research. This research successfully explores different aspects of temporal graph learning for dynamic recommendation, including temporal update, interaction scrutiny, and dynamic fairness. A promising direction for future work is to explore interpretability in temporal graphs by leveraging large language models to generate interpretable recommendations for users. This could further enhance the applicability and extensibility of this research. |
| Rights: | All rights reserved |
| Access: | open access |
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/200/14549