A smart data approach to traffic safety

Abstract

This work demonstrates, how a large data-base of traffic crashes can be used to analyze ensemble data. It fused data from the German Unfallatlas (German Crash Database - GCDB) with Open Streetmap (OSM) data (both publicly available), and a database from the German Federal State Northrhine-Westfalia (NW) named NWSIB that provides additional information about each intersection, most importantly an estimate of the ADT-values and street view images. The results have to be taken with care, since the quality of the ADT's in the database is hard to control, and because this approach may have assignment errors. The results partially reproduce known findings; however, they allow in principle for a more detailed investigation of the relationship between crash-numbers and ADT-values than is possible with generalized linear models (glm). In line with the call, all the data, as well as the scripts that analyse the data are publicly available -- this text is entirely written in the Rmd format \citep{RStudio2022}, and most computations have been done in R \citep{R2021} and QGIS \citep{QGIS}.