Dynamic Origin–Destination Matrix Prediction with Line Graph Neural Networks and Kalman Filter

Xi Xiong, Kaan Ozbay, Li Jin, Chen Feng

Research output: Contribution to journalArticlepeer-review


Modern intelligent transportation systems provide data that allow real-time dynamic demand prediction, which is essential for planning and operations. The main challenge of prediction of dynamic origin–destination (O-D) demand matrices is that demand cannot be directly measured by traffic sensors; instead, it has to be inferred from aggregate traffic flow data on traffic links. Specifically, spatial correlation, congestion and time dependent factors need to be considered in general transportation networks. This paper proposes a novel O-D prediction framework combining heterogeneous prediction in graph neural networks and Kalman filter to recognize spatial and temporal patterns simultaneously. The underlying road network topology is converted into a corresponding line graph in the newly designed fusion line graph convolutional networks (FL-GCNs), which provide a general framework of predicting spatial-temporal O-D flows from link information. Data from the New Jersey Turnpike network are used to evaluate the proposed model. The results show that the proposed approach yields the best performance under various prediction scenarios. In addition, the advantage of combining deep neural networks and Kalman filter is demonstrated.

Original languageEnglish (US)
Pages (from-to)491-503
Number of pages13
JournalTransportation Research Record
Issue number8
StatePublished - Jan 1 2020

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Mechanical Engineering


Dive into the research topics of 'Dynamic Origin–Destination Matrix Prediction with Line Graph Neural Networks and Kalman Filter'. Together they form a unique fingerprint.

Cite this