Bikesharing systems’ popularity has continuously been rising during the past years due to technological advancements. Managing and maintaining these emerging systems are indispensable parts of these systems and are necessary for their sustainable growth and successful implementation. One of the challenges that operators of these systems are facing is the uneven distribution of bikes due to users’ activities. These imbalances in the system can result in a lack of bikes or docks and consequently cause user dissatisfaction.

A dynamic repositioning model that integrates prediction and routing is proposed to address this challenge. This operational model includes prediction, optimization, and simulation modules and can assist the operators of these systems in maintaining an effective system during peak periods with less number of unmet demands. It also can provide insights for planners by preparing development plans with the ultimate goal of more efficient systems.

Developing a reliable prediction module that has the ability to predict future station-level demands can help system operators cope with the rebalancing needs more effectively. In this research, we utilize the expressive power of neural networks for predicting station-level demands (number of pick-ups and drop-offs) of bikeshare systems over multiple future time intervals. We examine the possibility of improving predictions by taking into account new sources of information about these systems, namely membership type and status of stations.

A mathematical formulation is then developed for repositioning the bikes in the system with the goal of minimizing the number of unmet demands. The proposed module is a dynamic multi-period model with a rolling horizon which accounts for demands in the future time intervals. The performance of the optimization module and its assumptions are evaluated using discrete event simulation. Also, a three-step heuristic method is developed for solving large-size problems in a reasonable time. Finally, the integrated model is tested on several case studies from Capital Bikeshare, the District of Columbia’s bikeshare program.