Spatiotemporal Demand Forecasting and Proactive Rebalancing Optimization for Urban Bike-Sharing: An Integrated Machine Learning Framework

This study presents an integrated, weather-aware framework for urban bike-sharing systems that combines Random Forest-based station-level demand forecasting with zone-level linear programming rebalancing optimization. Using 2024 Jersey City Citi Bike data, the model achieves robust predictive performance and a 13.4% improvement in service fulfillment rate alongside an 18.2% reduction in daily rebalancing vehicle trips compared to a reactive baseline, offering a practical, deployable solution for proactive fleet management.本研究提出了一款面向城市共享单车系统的整合式天气感知框架，该框架将基于随机森林（Random Forest）的站点级需求预测与区域级线性规划再平衡优化相结合。本研究采用2024年泽西城Citi Bike数据集开展实验，相较于被动式基准方案，该模型展现出优异的预测性能，服务履约率提升13.4%，每日再平衡调度车辆行程减少18.2%，可为主动式车队管理提供可落地的实用解决方案。