The demand balance for dockless bike-sharing systems (DBS) has become an important concern for governments and operators. Due to its lack of fixed sites, DBS is significant heterogeneity. Existing studies pay little attention to the impact of cycling environments on DBS heterogeneity. This study applies the random forest algorithm and a cluster-based approach to quantify the effects of cycling environments on the heterogeneity of bike-sharing trips. A probabilistic model considering cycling environments is also constructed to predict the bike-sharing flows among communities. A case study from Huangpu District in Shanghai shows that approximately 6% of trips are reduced for each 1% increase in gradient. Additionally, 50.79% of bike-sharing trips are made on commercial land. The performance of the proposed prediction model considering cycling environments outperforms the state of the art work, and its prediction accuracy is 0.76. The modeling framework provides valuable proposals for bicycle facilities planning and bicycle dispatch.

无桩共享单车系统 (DBS) 的需求平衡已成为政府和运营商的重要关注点。由于缺乏固定位点，DBS具有显着的异质性。现有研究很少关注循环环境对 DBS 异质性的影响。本研究应用随机森林算法和基于聚类的方法来量化骑行环境对自行车共享出行异质性的影响。还构建了一个考虑自行车环境的概率模型来预测社区之间的自行车共享流量。上海黄浦区的一项案例研究表明，坡度每增加 1%，出行量就会减少约 6%。此外，50.79%的共享单车出行发生在商业用地。所提出的考虑骑行环境的预测模型的性能优于现有技术，其预测精度为 0.76。该建模框架为自行车设施规划和自行车调度提供了有价值的建议。