学术报告20250530：Integrated Traffic Flow Prediction and Congestion Analysis: Exploring Physics-Informed Multi-View Clustering and Group Sparse Optimization

报告题目：Integrated Traffic Flow Prediction and Congestion Analysis: Exploring Physics-Informed Multi-View Clustering and Group Sparse Optimization

报告人：罗自炎

摘 要：In this talk, we attempt to incorporate the Physics-Informed Multi-View Clustering and the Group Sparse Optimization approach for traffic flow prediction and traffic congestion analysis. To analyze spatial heterogeneity, we utilize a multi-view structure that combines the Stepwise Common Principal Component with the k-means method by partitioning dataset into small ones. As a key component of the architecture, an advanced L2,0-regularized group sparse optimization with ridge regression loss is proposed for the underlying multivariate linear regression, aiming at reducing redundancy in historical traffic information by capturing the inherent temporal correlation in each subset. A time-efficient subspace Newton method is also applied to handle the proposed discontinuous and nonconvex tensor optimization problem. To further analyze traffic congestion, a physics-informed approach is adopted to show the relationship between the demand-to-capacity ratio and the congestion duration. We conduct experiments on five real datasets by comparing several state-of-the-art prediction methods, such as DCRNN, FGRU, PLS, etc. Numerical results demonstrate the efficiency of our proposed approach both in traffic flow prediction and congestion analysis.

报告人简介：罗自炎，北京交通大学数学与统计学院教授、博士生导师，中国运筹学会数学规划分会副秘书长，中国运筹学会女性工作委员会委员，中国运筹学会算法软件与应用分会理事。曾访问美国斯坦福大学、新加坡国立大学、香港理工大学、英国南安普顿大学等。主要从事张量优化、稀疏优化及统计优化的理论、算法及应用研究，在SIOPT、MP、MOR、IEEE TSP、JMLR等顶级期刊发表学术论文，合著SIAM出版社英文专著1部，曾获教育部自然科学奖二等奖、中国运筹学会青年科技奖提名奖、2023年入选国家级青年人才计划。

报告时间：2025年5月30日下午16:00-17:00

报告地点：内蒙古大学玉泉校区西院创新北楼512

欢迎广大师生参加！

内蒙古大学数学科学学院

2025年5月28日