Bikelution: Federated Gradient-Boosting for Scalable Shared Micro-Mobility Demand Forecasting
This work addresses privacy-aware demand forecasting for dockless bike-sharing systems, offering a scalable solution with incremental improvements over existing federated learning methods.
The paper tackled the problem of forecasting bike-sharing demand up to six hours ahead by addressing privacy and bandwidth issues in centralized approaches, proposing Bikelution, a federated learning solution using gradient-boosted trees that achieved comparable accuracy to centralized methods and outperformed state-of-the-art models on three real-world datasets.
The rapid growth of dockless bike-sharing systems has generated massive spatio-temporal datasets useful for fleet allocation, congestion reduction, and sustainable mobility. Bike demand, however, depends on several external factors, making traditional time-series models insufficient. Centralized Machine Learning (CML) yields high-accuracy forecasts but raises privacy and bandwidth issues when data are distributed across edge devices. To overcome these limitations, we propose Bikelution, an efficient Federated Learning (FL) solution based on gradient-boosted trees that preserves privacy while delivering accurate mid-term demand forecasts up to six hours ahead. Experiments on three real-world BSS datasets show that Bikelution is comparable to its CML-based variant and outperforms the current state-of-the-art. The results highlight the feasibility of privacy-aware demand forecasting and outline the trade-offs between FL and CML approaches.