Data-Driven Bike Sharing System Optimization - State of the Art and Future Opportunities.

The prevalence of public bike sharing systems (BSS) in recent years provides a last-mile-trip supplement to other modes of public transit systems in a convenient and environmentally-friendly way. However, due to the inherent imbalance between the renting and return rates, it is very challenging for the operators to optimize the resources (e.g., stations and bikes) to meet the fluctuating demand appropriately. Traditionally, urban planners use surveys to guide the BSS optimization, which is costly and inefficiently. With the availability of multi-source data in the city, it is now an emerging trend to leverage multi-source urban data for optimizing BSS. In this article, we specifically focus on the datadriven BSS optimization problem and provide a comprehensive literature review with future research opportunities. We first identify the main challenges and present a general technical framework for BSS optimization, and then present the representative studies in each sub-problem. Finally, we charter the research directions and opportunities in the future. Keywords Bike Sharing System Optimization; Demand Prediction; Station Placement; Bike Balancing 1