A data-driven dynamic repositioning model in bicycle-sharing systems

The new generation of bicycle-sharing is an O2O (online-to-offline) platform service that enables the users to access bicycle with a smartphone App. This paper proposes dynamic repositioning model predicted demand, where time interval fixed. A data-driven Neural Network (NN) approach introduced forecast demand. objective function at each defined simultaneously minimize operator cost and penalty cost. In addition normal constraints in static problem, flow conservation, inventory-balance travel are taken into account. Due non-deterministic polynomial-time hard (NP-hard) nature this model, hybrid metaheuristic Adaptive Genetic Algorithm (AGA) Granular Tabu Search (GTS) algorithm applied calculate solution. Based on initial plan made by AGA statically beginning study horizon, which ensures global optimization first As goes on, checked updated according real-usage patterns using GTS algorithm, has advantage high-performance local-search within short computing time. Numerical analysis conducted real cases. simulation results reveal proposed methodology can effectively problem response real-time usage. be value-added tool enhancing feasibility sustainability program.