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

Area of Science:

• Traffic Safety Research
• Data Science in Transportation
• Public Health

Background:

• Police reports on e-scooter incidents in Germany lack standardization and public accessibility.
• A need exists for data-driven analysis of e-scooter crashes and their contributing factors.

Purpose of the Study:

• To analyze German police reports on e-scooter crashes using web content mining and natural language processing.
• To identify key factors, consequences, and rider behaviors associated with e-scooter accidents.

Main Methods:

• Systematic web content mining of 1936 crash-related police reports from the German press portal over two years.
• Sentiment analysis to assess report objectivity, followed by keyword identification and network analysis (Louvain algorithm) to explore relationships.

Main Results:

• E-scooter crashes frequently involve driving under the influence of drugs (especially alcohol), riding with multiple passengers, and violating traffic rules (e.g., wrong-way riding).
• Consequences include severe injuries, license revocation, fines, criminal charges, and property damage.
• Low acceptance of protective gear and helmets among riders was observed.

Conclusions:

• Recommendations include nighttime e-scooter bans in certain areas, mandatory driving tests, and promoting helmet use.
• Addressing impaired driving and risky behaviors is crucial for improving e-scooter safety.