TY - JOUR
T1 - Dynamic demand-driven bike station clustering
AU - Wang, Yi-Jia
AU - Kuo, Yong-Hong
AU - Huang, George Q.
AU - Gu, Weihua
AU - Hu, Yaohua
N1 - Funding Information:
This research was partially supported by the Germany/Hong Kong Joint Research Scheme, Germany Academic Exchange Service (DAAD) and Research Grants Council (RGC) of Hong Kong ( G-HKU703/19 ), General Research Fund (GRF) , Research Grants Council (RGC) of Hong Kong (grant 17200820 ), Innovation and Technology Support Programme (Platform Project) (grant number ITP/021/20LP ), and the 2019 Guangdong Special Support Talent Program—Innovation and Entrepreneurship Leading Team (China) ( 2019BT02S593 ).
Publisher Copyright:
© 2022 Elsevier Ltd
PY - 2022/4/1
Y1 - 2022/4/1
N2 - As an eco-friendly transportation option, bike-sharing systems have become increasingly popular because of their low costs and contributions to reducing traffic congestion and emissions generated by vehicles. Due to the availability of bikes and the geographically varied bike flows, shared-bike operators have to reposition bikes throughout the day in a large and dynamic shared-bike network. Most of the existing studies cluster bike stations by their geographical locations to form smaller sub-networks for more efficient optimization of bike-repositioning operations. This study develops a new methodological framework with a demand-driven approach to clustering bike stations in bike-sharing systems. Our approach captures spatiotemporal patterns of user demands and can enhance the efficiency of bike-repositioning operations. A directed graph is constructed to represent the bike-sharing system, whose vertices are bike stations and arcs represent bike flows, weighted by the number of trips between the bike stations. A novel demand-driven algorithm based on community detection is developed to solve the clustering problem. Numerical experiments are conducted with the data captured from the world's largest bike-sharing system, consisting of nearly 3000 stations. The results show that, with CPLEX solutions as the benchmark, the proposed methodology provides high-quality solutions with shorter computing times. The clusters identified by our methodology are effective for bike repositioning, demonstrated by the balance of bike flows among clusters and geographic proximity of bike stations in each cluster The comparison between clusters found in different hours indicates that bike sharing is a short-distance transportation mode. One of the key conclusions from the computational study is that clustering bike stations by bike flow in the network not only enhances the efficiency of bike-repositioning operations but also preserves the geographic characteristics of clusters.
AB - As an eco-friendly transportation option, bike-sharing systems have become increasingly popular because of their low costs and contributions to reducing traffic congestion and emissions generated by vehicles. Due to the availability of bikes and the geographically varied bike flows, shared-bike operators have to reposition bikes throughout the day in a large and dynamic shared-bike network. Most of the existing studies cluster bike stations by their geographical locations to form smaller sub-networks for more efficient optimization of bike-repositioning operations. This study develops a new methodological framework with a demand-driven approach to clustering bike stations in bike-sharing systems. Our approach captures spatiotemporal patterns of user demands and can enhance the efficiency of bike-repositioning operations. A directed graph is constructed to represent the bike-sharing system, whose vertices are bike stations and arcs represent bike flows, weighted by the number of trips between the bike stations. A novel demand-driven algorithm based on community detection is developed to solve the clustering problem. Numerical experiments are conducted with the data captured from the world's largest bike-sharing system, consisting of nearly 3000 stations. The results show that, with CPLEX solutions as the benchmark, the proposed methodology provides high-quality solutions with shorter computing times. The clusters identified by our methodology are effective for bike repositioning, demonstrated by the balance of bike flows among clusters and geographic proximity of bike stations in each cluster The comparison between clusters found in different hours indicates that bike sharing is a short-distance transportation mode. One of the key conclusions from the computational study is that clustering bike stations by bike flow in the network not only enhances the efficiency of bike-repositioning operations but also preserves the geographic characteristics of clusters.
KW - Bike repositioning
KW - Bike sharing
KW - Community detection
KW - Demand-driven clustering
UR - http://www.scopus.com/inward/record.url?scp=85126289222&partnerID=8YFLogxK
U2 - 10.1016/j.tre.2022.102656
DO - 10.1016/j.tre.2022.102656
M3 - Journal article
SN - 1366-5545
VL - 160
JO - Transportation Research, Part E: Logistics and Transportation Review
JF - Transportation Research, Part E: Logistics and Transportation Review
IS - 102656
M1 - 102656
ER -