Expanding a (electric) bicycle-sharing system to a new city: Prediction of demand with spatial regression and random forests

Abstract

The number of bicycle-sharing systems has undergone strong growth in the last two decades. This growth is part of a worldwide trend that has began in the 1990s and has strongly accelerated after 2005. Early bicycle-sharing systems have mainly been provided as a public service by cities, but meanwhile major international bicycle-sharing companies have emerged that seek to expand their operations to new cities. Two major strategic questions that arise are which cities should be considered for an expansion and the geographical extent of the service area. An important factor to support these decisions is expected demand for bicycle-sharing, as it is directly related to potential revenue. In this paper, booking data from an electric bicycle-sharing system was used to estimate and assess models for bicycle-sharing demand and to make predictions for an expansion to a new city. Employment, population, bars and restaurants and distance to a central location were among the most important predictors in terms of variance explained in the same city. However, omitting centrality measures improved predictions for the new city.