Kai Yang(杨凯)

PhD Candidate

Harbin Institute of Technology (Shenzhen)
Supervisor: Prof. Xun Zhou
IEEE & ACM Student Member

Research interests: Spatial-temporal Data Mining, Recommendation System, Trustworthy AI

Email: kaiyang.cs AT outlook.com
[Google Scholar] [GitHub]

Kai Yang

Short Bio

I am currently a PhD student at Harbin Institute of Technology, Shenzhen (HITSZ), supervised by Prof. Xun Zhou. My research focuses on spatial-temporal data mining, recommendation system and trustworthy AI.

Before that, I received my B.Eng. and M.Eng. degrees from the School of Information and Software Engineering, University of Electronic Science and Technology of China (UESTC) in 2022 and 2025, respectively.

Recent News

[Show more]

Selected Publications

DecField
DecField: Learning Disentangled Electrostatic Fields for Human Spatiotemporal Decision-Making Prediction
Xinyue Wang, Kai Yang, Yajun Pan; Xun Zhou
KDD 2027, CCF-A

DecField proposes an ante‑hoc human spatiotemporal decision‑making framework built on disentangled electrostatic energy‑field learning, which decouples environmental impacts from personal preferences while delivering both competitive prediction performance and intuitive field‑visualized explanations.

SHAPE
SHAPE: Spatial Hyperbolic-Embedding Augmented Pretraining for Estimated Time of Arrival on Obfuscated Trajectories
Ruopu Li, Yong Chu, Kai Yang, James Jianqiao Yu, Lixia Wu, Xun Zhou
ICDM 2026, CCF-B

SHAPE is a spatial hyperbolic-embedding augmented pretraining framework for estimated time of arrival on obfuscated trajectories, which learns robust route representations via differential geometric feature decomposition and hybrid Euclidean-hyperbolic contrastive self-supervised learning.

RoPOI
RoPOI: Robust POI Recommendation based on Modality Disentanglement and Missing Representation Generation
Kai Yang, Xun Zhou
ACM MM 2026, CCF-A

RoPOI proposes a robust point-of-interest recommendation framework built on modality disentanglement and missing representation generation, which mitigates performance degradation from modality absence while preserving discriminative modality-specific information.

ExNext
ExNext: Self-Explainable Next POI Recommendation [Paper] [Code]
Kai Yang, Yi Yang, Qiang Gao, Ting Zhong, Yong Wang, Fan Zhou
SIGIR 2024, CCF-A

ExNext proposes a self-explainable POI recommendation framework that improves trustworthiness by balancing accuracy and explainability, leveraging information theory to learn compact, informative representations.

ExGeo
ExGeo: Exploring Self-Explainable Street-Level IP Geolocation with Graph Information Bottleneck [Paper] [Code]
Kai Yang, Wenxin Tai, Zhenhui Li, Ting Zhong, Guangqiang Yin, Yong Wang, Fan Zhou
ICASSP 2024, CCF-B

ExGeo endows the IP geolocation model with explainability via a variational graph information bottleneck, balancing concise explanation and informative representation.

[Show more]

Honors

Services