Exploring the determinants of hydrogen bike-sharing adoption among urban residents using explainable machine learning

Hydrogen bike-sharing is an emerging mode of zero-carbon transport that has recently appeared in some cities. However, salient empirical research remains insufficient, and it is thus unclear how to effectively increase residents’ adoption intention. This study draws on innovation diffusion theory (IDT) to develop a model of the determinants of hydrogen bike-sharing adoption intention, including innovation attributes, social system, communication channels, time, and external environment characteristics. Using data collected from 5,220 residents across 302 cities in China, we employed LightGBM to train and test the model. The results of this model combined with Shapley additive explanations (SHAP) are threefold. (i) In terms of relative importance, the relative advantage of innovation, social system, communication channels, and the external environment are important, whereas the importance of time is limited. (ii) Most determinants exhibit pronounced nonlinear and threshold effects. Particularly, price and maximum riding range reveal a diminishing marginal effect of relative advantage, with adoption intention increasing once they outperform conventional bike-sharing options but rising more slowly as the advantage further expands. (iii) Interaction effects between external environmental characteristics and key IDT factors are observed. The role of price as an economic relative advantage varies with the level of hydrogen infrastructure, while the performance relative advantage of maximum riding range varies across urban areas. These findings provide decision makers and practitioners with more targeted strategies for accelerating the transition toward zero-carbon transport.