Export
Haina Tang
2 previous affiliations
- Associate Professor · 2015–2026
- Sun Yat-sen University2008–2012
- PhD student · 2008–2012
Research Overview
Haina Tang focuses on spatio-temporal data mining, trajectory analysis, and intelligent transportation systems. Early work addressed trajectory-user linking and similarity computation using deep learning techniques to handle noisy, real-world location data. More recently, Tang's research has explored predictive modeling and reinforcement learning for optimizing ride-hailing and taxi demand distribution.
Research Trajectory
Tang’s published research trajectory begins in the domain of spatio-temporal data mining, with a specific focus on extracting reliable signals from complex, noisy trajectory data. In 2020, Tang contributed as second author to TULSN: Siamese Network for Trajectory-user Linking. This work addressed the challenge of linking trajectories to their owners in location-based social networks—a critical task for personalized recommendation and anomaly detection. The paper highlighted the limitations of traditional rule-based methods, which often ignore sequence semantics, and standard learning-based methods, which require extensive retraining when new users are introduced. By employing a Siamese network architecture, the research provided a more scalable approach to mapping trajectories to users.
Continuing this focus on foundational trajectory analysis, Tang was a second author on On Accurate Computation of Trajectory Similarity via Single Image Super-Resolution in 2021. Measuring trajectory similarity is often complicated by real-world data quality issues, such as non-uniform sampling rates and sensor noise. Traditional pairwise point-matching methods struggle with inconsistent sampling, while recurrent neural network (RNN) approaches can introduce compounding errors by conditioning predictions on their own generated historical data. This research proposed an alternative framing, adapting single image super-resolution techniques to complement and refine trajectory data for more accurate similarity measurements.
By 2022, Tang’s collaborative work expanded from fundamental trajectory processing to applied intelligent transportation systems, participating as a co-author in broader research teams. Tang was a contributing middle author on HMDRL: Hierarchical Mixed Deep Reinforcement Learning to Balance Vehicle Supply and Demand, which tackled the persistent problem of supply-demand imbalance in online ride-hailing networks. The paper introduced a three-level hierarchical mixed deep reinforcement learning framework designed to reposition idle vehicles efficiently, utilizing a top-level manager for time-dimension abstraction, middle-level coordinators, and bottom-level executive workers.
In the same year, Tang also collaborated on Taxi Demand Prediction Using Parallel Multi-Task Learning Model. This research addressed the need for real-time taxi resource allocation by co-predicting both pick-up and drop-off demands. Rather than treating these as isolated spatial-temporal problems, the team utilized a multi-task learning model with parallel LSTM layers to capture the inherent correlations between the two demand types, demonstrating improved performance over single-demand prediction methodologies.
Current Directions
Based on the most recent papers in the evidence from 2022, Tang’s research interests have moved toward the optimization of urban mobility networks and fleet management. While earlier work concentrated heavily on the algorithmic challenges of processing individual trajectory sequences, recent collaborative efforts investigate macro-level transportation dynamics. This includes the application of multi-task learning architectures to forecast spatial-temporal demand patterns and the use of hierarchical reinforcement learning to automate the repositioning of ride-hailing fleets. These directions reflect a broader interest in applying advanced neural network architectures to solve real-time resource allocation and operational efficiency problems in modern smart city infrastructures.
Profile updated
Loading new papers…
Loading research neighbors…
See something wrong? Report a mistake