Data-Driven Optimisation of Supply-Demand Balance for Shared Bicycles

With the rapid proliferation of shared bicycles in urban transport, the imbalance between supply and demand distribution has become increasingly pronounced. The coexistence of vehicle shortages in high-demand areas and idle bicycles in low-demand zones is particularly evident during morning and evening rush hours. Existing research typically treats hotspot identification, demand forecasting, and dispatch optimisation as independent tasks, hindering the formation of a unified decision chain and constraining improvements in dispatch efficiency. To address this, this paper constructs an operational optimisation framework integrating spatial clustering, demand forecasting, and intelligent dispatch. First, irregularly distributed hotspot areas are identified through grid processing and DBSCAN density clustering. Second, a high-precision short-term demand forecast is achieved using an XGBoost model that integrates spatio-temporal features with historical data. Finally, based on the forecast results, an ant colony optimisation algorithm generates dispatch plans that balance both path efficiency and utilisation rates. Experiments using five days of real-world cycling data from Xiamen demonstrate that DBSCAN captures hotspot structures more accurately than K-Means, XGBoost achieves a 34% reduction in RMSE compared to the baseline model, and the ant colony algorithm reduces dispatch distances by 18% relative to a greedy strategy. These findings validate the framework's effectiveness in mitigating supply-demand imbalances and enhancing operational efficiency.