Efficiency Comparison of Public Bike-Sharing Repositioning Strategies Based on Predicted Demand Patterns

Abstract

Public bike-sharing systems are used worldwide, and the imbalance between supply and demand for bicycles and operational inefficiency is becoming increasingly severe. For a system to operate efficiently, it is necessary to relocate bicycles among rental stations to minimize a lack of bikes at the station causing unmet demand. Recent studies have presented various repositioning strategies for bike-sharing systems and compared their efficiency. However, little consideration has been paid to the strategy of the spatial and temporal patterns of bike-sharing demand and the inventory level. This study aims to analyze the spatiotemporal patterns of the forecasted demand for the bike-sharing system and to compare the efficiency of different repositioning strategies to choose the most efficient one. We use three repositioning strategies with different additional constraints related to unbalanced stations and present computational results with real data in Seoul. Two indices represent the temporal variation of predicted inventory at each station and the coefficients of the spatial variation for hourly unmet demand. Linear classifiers are derived by linear discriminant analysis to classify the efficiency of each strategy according to developed indices. The study reveals that adding constraints of imbalanced stations to the strategy according to the spatiotemporal characteristics of forecasted inventory can help to reduce unmet demand. The result of this study enables proactive decision-making using proposed indices in operating bike-sharing systems and contributes to improving the efficiency and reliability of systems.