Koliou P., Theodoraki E. M., Yannis G., “From Trend Observation to Risk Interpretation: A Machine Learning Analysis of Micromobility Injury Severity in 2022”, Proceedings of the Transport Research Arena TRA 2026 Conference, Budapest, 18-21 May 2026.

Longitudinal analysis of micromobility-related traffic collisions across Europe (2013–2023) shows a steady increase, with 2022 marking a pronounced spike across all severity categories. This study conducts a severity-risk assessment for 2022, identifying key factors influencing serious injury outcomes. Using multi-country accident data, we trained Random Forest models distinguishing between micromobility users and other vehicles. Variables included demographics, environmental conditions, lighting, and spatial context. Interpretability was ensured with SHapley Additive exPlanations (SHAP). Results show that for micromobility users, age is the strongest predictor, with heightened risk among younger (18–24) and older (65+) riders. Weather, lighting, and rural settings also contributed. For other users, rural context, weather, and male gender dominated. Models achieved moderate predictive performance (63.1% and 67.2%), with SHAP providing clear insights. This approach highlights differentiated risk profiles and supports targeted, evidence-based safety interventions.