Hierarchical Prediction Based on Two-Level Affinity Propagation Clustering for Bike-Sharing System

• Published in: IEEE access Vol. 6; pp. 45875 - 45885
• Main Authors: Jia, Wenzhen, Tan, Yanyan, Li, Jing
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Affinity; affinity propagation clustering; Algorithms; Bicycles; Bicycling; Bike-sharing system; Clustering; Clustering algorithms; Forecasting; Geographical locations; Mathematical models; Meteorology; migration trend; prediction; Prediction algorithms; Prediction models; Predictive models; Propagation; Stations
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2018.2865658

Bike-sharing system is a new transportation that has emerged in recent years. More and more people will choose to ride bicycle sharing at home and abroad. While we use shared bicycles conveniently, there are also unfavorable factors that affect the customer's riding experience in the bicycle-sharing system. Due to the rents or returns of bikes at different stations in different periods are imbalanced, the bikes in the system need to be rebalanced frequently. Therefore, there is an urgent need to predict and reallocate the bikes in advance. In this paper, we propose a hierarchical forecasting model that predicts the number of rents or returns to each station cluster in a future period to achieve redistribution. First, we propose a two-level affinity propagation clustering algorithm to divide bike stations into groups where migration trends of bikes among stations as well as geographical locations information are considered. Based on the two-level hierarchy of stations, the total rents of bikes are predicted. Then, we use a multi-similarity-based inference model to forecast the migration proportion of inter-cluster and across cluster, based on which the rents or returns of bikes at each station can be deduced. In order to verify the effectiveness of our two-level hierarchical prediction model, we validate it on the bike-sharing system of New York City and compare the results with those of other popular methods obtained. Experimental results demonstrate the superiority over other methods.