A Study on the Integration of Demand Forecasting and Optimization Models for Enhancing the Operational Efficiency of Shared Bicycle Systems

The aim of this paper is to develop a framework for optimizing the daily redistribution of shared bicycles by integrating demand forecasting and mathematical optimization techniques. This framework addresses the operational challenges associated with balancing supply-demand dynamics across bicycle stations while minimizing costs. First, several potential demand forecasting models are constructed by combining time-series algorithms, such as Prophet and SARIMAX, in a stacking framework to predict rental and return demands. Second, an optimization model for bicycle redistribution is formulated as a capacitated vehicle routing problem with pickups and deliveries to minimize redistribution costs and improve system efficiency. Third, the significance of the demand forecasting models is evaluated based on their predictive accuracy and operational impact, allowing prioritization of effective forecasting and optimization strategies. Finally, a stepwise decision-making process for daily redistribution is established based on the prioritized results. We illustrate the proposed framework using real-life data of shared bicycle systems in metropolitan areas, showcasing its potential to enhance operational efficiency and decision-making for urban mobility services.