Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Political philosophy. Social philosophy --- Logic

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Crowdsourced traffic data is increasingly available in this big data world. Governments, traffic managers and traffic institutions are aware of this trend and wonder whether it can complement convential data sources and provide additional knowledge about traffic situations. This study focuses on traffic safety in particular by investigating crowdsourced Waze accident alerts and comparing them to official police crash reports. The aim is to gain insight in spatial coverage and temporal accuracy of Waze data and examine whether it can provide additional value to crash statistics compared to police crash data. oindent Ten weeks of crash data in the province of Limburg in Belgium is analyzed for this study. The findings indicated that Waze reports have reasonable accuracy in space (on average 129 meters around police report) and time (on average 12 minutes after the presumed timestamp in the police report). After a reliability assessment and a regression model on the matching between Waze and police data, this study suggests to consider Waze alerts that have a reportRating of 1 or more to be more valid than other alerts. oindent After matching Waze accident alerts to police reports, the overlap turned out to be quite low (13.6% of police reports matched with 23.9% of Waze reports), which indicates that both datasets likely suffer from serious underreporting. This study estimates a crash count for the considered case study of 5490-6136 crashes using capture-recapture method on the stratified datasets. Police reports capture 22-25% of this estimated actual number of traffic crashes during the considered period and region, Waze captures 15-16%. In addition to these findings, a procedure to evaluate crowdsourced data is presented. Using this procedure, the additional coverage of Waze in the estimated crash count can be calculated. Without Waze and considering that the police publishes all its data (i.e. also PDO crash data), there is knowledge about 22.3-25.0% of crashes. Including Waze information increases this percentage to 33.6-37.5%, implying that Waze is a valuable source of information supplementary to police data.