TY - JOUR

T1 - Web content mining analysis of e-scooter crash causes and implications in Germany

AU - Brauner, Tim

AU - Heumann, Maximilian

AU - Kraschewski, Tobias

AU - Prahlow, Oliver

AU - Rehse, Jan

AU - Kiehne, Christian

AU - Breitner, Michael H.

N1 - Funding Information: ☆ Acknowledgments This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

PY - 2022/12

Y1 - 2022/12

N2 - In Germany, police reports published via press are neither uniformly written nor accessible to the public. There is a lack of comprehensive and factual data-based analyses of e-scooter crashes and their causes. We collected 1936 crash-related reports over two years via the German press portal based on a systematic web content mining process. Sentiment analysis results revealed that the police reports’ coverage is predominantly factual and neutral and, therefore, useful for keyword-based analyses. After identifying the 46 most relevant keywords in the reports, we generated an adjacency matrix to investigate the keywords’ dependencies, visualized the network and dependencies of the most relevant keywords, and categorized them into four thematic clusters using the Louvain algorithm. Our results and findings reveal that driving under drug influence, especially alcohol, is one serious problem. Riding e-scooter in pairs and on forbidden terrain or in the wrong direction are also common causes of crashes. Consequences for e-scooter riders are severe injuries, driving license revocation, fines, criminal charges, and incurring for property damage. Further, wearing protective gear and helmets is of low acceptance among the e-scooter ridership. Based on our results and findings, we recommend e-scooter bans during the night times for some locations, obligatory driving tests before first e-scooter use, and helmet wearing.

AB - In Germany, police reports published via press are neither uniformly written nor accessible to the public. There is a lack of comprehensive and factual data-based analyses of e-scooter crashes and their causes. We collected 1936 crash-related reports over two years via the German press portal based on a systematic web content mining process. Sentiment analysis results revealed that the police reports’ coverage is predominantly factual and neutral and, therefore, useful for keyword-based analyses. After identifying the 46 most relevant keywords in the reports, we generated an adjacency matrix to investigate the keywords’ dependencies, visualized the network and dependencies of the most relevant keywords, and categorized them into four thematic clusters using the Louvain algorithm. Our results and findings reveal that driving under drug influence, especially alcohol, is one serious problem. Riding e-scooter in pairs and on forbidden terrain or in the wrong direction are also common causes of crashes. Consequences for e-scooter riders are severe injuries, driving license revocation, fines, criminal charges, and incurring for property damage. Further, wearing protective gear and helmets is of low acceptance among the e-scooter ridership. Based on our results and findings, we recommend e-scooter bans during the night times for some locations, obligatory driving tests before first e-scooter use, and helmet wearing.

KW - Accident analysis

KW - E-scooter

KW - Network graph

KW - Sentiment analysis

KW - Web content mining

UR - http://www.scopus.com/inward/record.url?scp=85138818592&partnerID=8YFLogxK

U2 - 10.1016/j.aap.2022.106833

DO - 10.1016/j.aap.2022.106833

M3 - Article

AN - SCOPUS:85138818592

VL - 178

JO - Accident Analysis and Prevention

JF - Accident Analysis and Prevention

SN - 0001-4575

M1 - 106833

ER -