Data-driven adaptive threshold control for bike share systems

Abstract

When it comes to building a successful public bike share such as the NYC CitiBike system, there are many important questions to answer. There is no shortage of work being done on finding the most efficient way to redistribute bikes to stations. Equally important to how to distribute bikes is the question of when to redistribute bikes. Redistribute too infrequently and customers become frustrated, resulting in decreased revenue. Redistribute too frequently and the cost of redistribution becomes prohibitively high. In this piece of research, we attempt to find the optimal time to call for the redistribution of bikes to minimize cost and retain maximum membership.