Spatial Pattern Mining Technique for Identifying Urban Traffic Congestion Hotspots Using GPS Trajectory Data

Autonomous mobility, route planning, emergency response, and intelligent transportation system (ITS) have been impacted by urban traffic congestion. Road-network connectivity and recurring congested relationships have been considered in existing GPS-based methods such as point-based clustering, density estimation, and spatial analysis, causing hotspot detection to be fragmented or unstable. This paper has proposed an Adaptive Spatial Trajectory Pattern Mining (ASTPM) technique to discover connected and persistent congestion hotspots from GPS trajectory data. The method has been applied to the trajectory preprocessing, segmentation, road-network map matching, and Spatio-Temporal feature extraction based on speed, density, travel-time deviation and temporal recurrence. Congested road segments have been identified using the Adaptive Congestion Intensity (ACI) model and patterns of connected congestion have been mined using Spatial Connectivity-Aware Pattern Mining (SC-PM) algorithm using spatial adjacency, cooccurrence, support and persistence. Intensity, spatial support, persistence and network connectivity have been subsequently evaluated with hotspot validation. ASTPM has been able to attain a HSI of 91.8% with 93.2% spatial consistency, 90.7% temporal persistence, and 92.1% network connectivity, thus showing stable hotspot identification. The network-aware congestion hotspot detection has been provided in the proposed framework. Future research is planned to investigate real time trajectories, multimodal data, online learning and cross city validation.