Led by Prof Jia Weijia, deputy director of the lab, the team studied various cutting-edge machine learning methods for mining spatial-temporal information. They found that most of these methods only focus on predicting the results of a certain time step in the future, but to achieve reliable long-term prediction, it is imperative to filter out noise and prevent errors from spreading through complex correlations. Finally, the team proposed a method with high prediction accuracy. Compared with previous methods, this new method can significantly improve the accuracy of the prediction results through extensive evaluation of multiple real-world data sets.
The first author of the paper, Stark Lin Haoxing, is a second-year master’s student of computer science in the Faculty of Science and Technology. He joined the state key lab in September 2018 and has since been working under the guidance of Prof Jia. Lin mainly studies the prediction of spatial-temporal phenomena in urban-wise systems.