KDD 2025 was successfully held from August 3rd to 7th, 2025, in Toronto, Canada. The paper titled "Dynamic Localisation of Spatial-Temporal Graph Neural Network" by the research team led by Teacher Duan Wenying from the School of Artificial Intelligence at Nanchang University was successfully accepted into the Research track. This paper is also the only research paper from Nanchang University accepted by KDD this year. The paper was presented through an oral report and a poster presentation at the conference. The paper proposes DynAGS, a sparsification framework for adaptive spatial-temporal graph neural networks, aiming to maximize efficiency and accuracy in distributed deployment. This framework integrates dynamic localization, temporally evolving spatial graphs, and personalized localization, all orchestrated around a dynamic graph generator—a lightweight central module that utilizes cross-attention. The central module can integrate historical information independently of nodes to enhance the feature representation of nodes at the current moment. This improved feature representation is then used to generate dynamic sparse graphs without costly data exchange and supports personalized sparsification. The first author of the paper is Duan Wenying, a young faculty member of the School of Artificial Intelligence. The second author is undergraduate student Guo Shujun (Class of 2022). Assistant Professor He Xiaoxi from the University of Macau and Professor Rao Hong, Dean of the School of Software, are the co-corresponding authors. Assistant Professor Zhou Zimu from City University of Hong Kong and Professor Huang Wei, Vice President of Yichun University, are co-authors.

KDD (ACM SIGKDD Conference on Knowledge Discovery and Data Mining) is a CCF-A category international top-tier academic conference in the field of data mining, with an acceptance rate consistently below 20%. It is organized by the Association for Computing Machinery (ACM) and has been held for 30 editions to date.