Mapping micro-mobility risk: AI-powered geospatial analysis and predictive modelling☆

Micro-mobility vehicles have rapidly become widespread as a sustainable and practical alternative for urban transportation in recent years. In this study, micro-mobility vehicles refer to traditional bicycles, electric bicycles, and electric scooters, which represent the main categories of such modes involved in traffic crashes in T & uuml;rkiye. Despite their growing popularity, the safety implications of these vehicles have not yet been fully understood, and comprehensive research addressing crash patterns and associated risk factors is required. To this end, this study employs an artificial intelligence-driven geospatial and statistical methodology. Crash reports involving micro-mobility vehicles in T & uuml;rkiye between 2015 and 2023 were analysed. Seventeen independent variables and 102 sub-variables were identified and integrated into a GIS environment for spatial analysis. The impact levels of risk factors were assessed using six different Large Language Models (DeepSeek, GEMINI, Perplexity, ChatGPT, Copilot, and Poe). Crash risk maps and corresponding weight values were combined to produce an crash suitability map indicating the potential risk of micro-mobility crashes. Furthermore, the significance of these factors across different collision types was tested using a multinomial logistic regression model. To the best of the authors' knowledge, this is the first study to apply a macro-scale dataset and an AI-enhanced geospatial decisionmaking approach to analyse micro-mobility crashes. The findings highlight the need for local governments and urban planners to implement targeted safety measures in regions with high crash potential.

Journal

SAFETY SCIENCE

Year

2026

Volume

196

ISSN

0925-7535

eISSN

1879-1042

DOI

10.1016/j.ssci.2025.107107

WoS UID

WOS:001657876900001