Demand estimation of public bike-sharing system based on temporal and spatial correlation

Nowadays, public Bike-Sharing Systems (BSSs) are broadly deployed in many cities around the world. It is important to obtain accurate user demand of BSS for better system planning and bicycle scheduling. The actual user demand includes not only the users who are served, but also those who are not served by BSS. In this study, we take into account the situations that users are not served for the first time. We propose a three-step demand estimation model to infer the situations that users are not served from both the temporal and spatial correlation, based on the two characteristics of station usage, long-term stability and shortterm volatility. The demand estimation model proposed is evaluated based on Washington D.C. bike-sharing system and uses the comprehensive information of three datasets, user trip data, station status data, and station location data. Compared with the ground truth of user demand, the minimum relative error in the experimental results of the entire system is 45.5%.