Bicycle choice: machine learning approach to understanding university students’ attitudes toward cycling

• Daily transport mode is the strongest predictor of micro-mobility preferences among university students. • Random Forest model achieved 78.3 % accuracy in predicting bike vs. e-bike preferences. • Clustering analysis revealed three distinct behavioral segments with unique mobility traits. • Gender and car ownership significantly influence safety perception and adoption willingness. • Study offers policy insights to promote micro-mobility in resource-constrained urban contexts. This study investigates the factors influencing university students’ preferences for micro-mobility solutions, with a specific focus on conventional bicycles and electric bikes (e-bikes) in developing country context. Despite global efforts to promote cycling, micro-mobility adoption remains low in regions like Palestine due to safety concerns, economic barriers, and lack of infrastructure. To address this gap, we employ machine learning techniques, Random Forest and k-means clustering, to analyze survey data from 1,061 students at An-Najah National University. The analysis highlighted the role of factors such as gender, car ownership, daily transport mode in shaping preferences, safety perceptions, and willingness to adopt micro-mobility options. The Random Forest model provided insights into the most influential variables, while the clustering analysis segmented individuals into distinct groups, allowing for a more tailored understanding of micro-mobility behavior. The results showed that the daily transport mode is the most significant factor affecting safety perceptions and preference. Gender and car ownership also emerged as important factors, influencing willingness to adopt micro-mobility and preferences between bicycles and e-bikes. Sensitivity analysis was employed to evaluate the robustness of the models by measuring the impact of small changes in key variables on predictions. The models were evaluated using accuracy, feature importance, and sensitivity analysis. Random Forest achieved an accuracy of 78.3% in predicting preferences, highlighting daily mode choice as the most influential variable. The results offer practical insights for policymakers and urban planners, particularly in developing countries like Palestine, where economic and infrastructural challenges affect the adoption of micro-mobility solutions.