Road Traffic Anomaly Detection via Collaborative Path Inference from GPS Snippets.

Area of Science:

• Urban Computing
• Transportation Science
• Data Mining

Background:

• Detecting road traffic anomalies from GPS data is vital for urban computing as anomalies often indicate significant events.
• Noisy and sparse GPS data presents challenges for accurate anomaly detection.

Purpose of the Study:

• To propose a robust two-stage solution for detecting road traffic anomalies from GPS snippets.
• To address the challenges posed by noisy and sparse GPS data.

Main Methods:

• A Collaborative Path Inference (CPI) model using Conditional Random Field (CRF) with static and dynamic features.
• Dynamic context features learned collaboratively via tensor decomposition.
• A Road Anomaly Test (RAT) model to calculate anomaly degrees from inferred fine-grained trajectories.

Main Results:

• The proposed method was evaluated on a large-scale real-world dataset of taxi GPS data in Beijing.
• Demonstrated superior performance compared to baseline techniques in detecting road traffic anomalies.

Conclusions:

• The two-stage solution effectively addresses the challenges of noisy GPS data for traffic anomaly detection.
• The method offers advantages for urban computing applications requiring accurate traffic monitoring.