The ADAIT team joined VicRoads in 2013 to study the congestion pattern in and around Melbourne CBD. This part of the project aims to develop an efficient tool that discovers significant congestion propagation patterns. Frequent patterns of the congestion propagations may indicate the bottlenecks of the road network.
This system uses advanced web technologies and state-of-the-art machine learning (ML) algorithms. The live web interface is based on NICTA Subspace and 3D Cesium Bing map to provide a spatial and temporal display of tweets that are potentially related to transport issues. The crawled tweets are first filtered to show incidents in Australia, and then divided into different groups by online clustering and classification algorithms.
- Identify congested road segments: a segment is considered as congested at a specific snapshot if its travel time is longer than some threshold (based on its average travel time)
- Build causal congestion propagation patterns by analysing spatial and temporal correlation in traffic flows
- Discover the most frequent causal propagation trees: finds all item sets whose confidence of occurrence exceed a certain threshold; these frequent trees reveal recurrent congestion propagations and suggest inherent problems in existing road networks
The causal inference tool is capable to assist transport authorities and planners in:
- Congestion hotspot identification
- Impact propagation and network sensitivity analysis