Traffic forecasting approaches are critical to develop adaptive strategies for mobility. Near-term traffic forecasting is a foundational component of these strategies. Traffic patterns have complex spatial and temporal dependencies that make accurate forecasting on large highway networks a challenging task for machine learning models.
We present a deep learning approach for large scale traffic forecasting using graph-partitioning-based diffusion convolutional recurrent neural networks (DCRNNs). This approach uses a graph-partitioning method to decompose a large highway network into smaller networks and trains them independently. We demonstrate the efficacy of the graph-partitioning-based DCRNN approach to model the traffic on a large California highway network with 11,160 sensor locations.
About the Speaker: Tanwi Mallick is a postdoctoral appointee in Mathematics and Computer Science Division at Argonne National Laboratory since October 2018. Her works include the development of scalable data-efficient machine learning methods for network congestion modeling, high-performance computing, and traffic modeling. Prior to Argonne, she worked as a Senior Data Scientist at General Electric. She received her PhD in Computer Science from Indian Institute of Technology, Kharagpur, India.