Comparison of Waze crash and disabled vehicle records with video ground truth

• Main Authors: Noah Goodall, Eun Lee
• Format: Article
• Language: English
• Published: Elsevier 2019-06-01
• Series: Transportation Research Interdisciplinary Perspectives
• Subjects: Crowdsourcing; Incident data; Validation
• Online Access: http://www.sciencedirect.com/science/article/pii/S2590198219300193

Waze is a popular mobile phone navigation application that allows users to report incidents on roadways in real time. Over 560 government agencies have access to Waze reports, and many are using them as data sources for operations. This study evaluated the accuracy of Waze crash and disabled vehicle reports along a 2.7-mile section of urban freeway by comparing Waze reports with images from four traffic cameras. Because the cameras were pan-tilt-zoom capable and operated by transportation management center staff, a surrogate measure for transportation agency awareness was used, defined as the time at which the camera panned or zoomed to an ongoing incident. Of 40 crashes reported in Waze, 13 (33%) were confirmed primary reports and 2 (5%) were false alarms. Of the 560 disabled vehicle reports, 125 (22%) were confirmed primary reports and 131 (23%) were false alarms. For disabled vehicles, neither a Waze report's reliability score nor an incident's duration was correlated with report accuracy. For both crashes and disabled vehicles, transportation management center staff was usually aware of the incident before the first Waze report, although this may be biased as the study corridor had dense camera coverage. This is the first study to evaluate the accuracy of individual Waze crash and disabled vehicle reports using ground truth evidence.