基於CPSS之共享自行車存量調度分析系統

Abstract

為降低環境汙染、減少資源消耗，綠色運輸導向發展（Greening Transport Oriented Development）成為全球重視議題。然而近年因為受COVID-19疫情影響，各國開始重視社交距離，進而改變民眾搭乘公共運輸之習慣。共享自行車系統（Public Bicycle System，PBS）即為其一快速發展之綠色交通方式，為被動、機動之個人出行提供有效的替代方案。隨著共享自行車普及，產生供需失衡問題，因此共享自行車存量平衡成為重要課題。 本研究以台北市區為範圍，對其微笑單車（YouBike）之共享自行車系統服務，執行租賃分析與評估，並重新平衡租賃站之庫存車量。研究分為三部分，租賃需求預測、站點庫存車量風險、混和調度建議。第一部分：租賃需求預測，將計算租賃站最佳庫存車量，首先使各租賃站不同時段之用戶租借和歸還資料建模，並加入CPSS（Cyber-Physical-Social Systems）概念之社會化因子，將氣候、季節、假日等因素考量進去，分析各租賃站用戶租賃需求。統整後透過XGBoost（eXtreme Gradient Boosting）演算法來預測未來租借和歸還率。第二部分：站點庫存車量風險，則是將管理學科之風險水準理念納入研究，擬定運補需求指標，透過數值化之指標將缺車和缺位站點歸納至熱點區。第三部分：混和調度建議，將透過預測分析出的租賃站最佳庫存車量與熱點區位，將運補策略劃分為動態與靜態調度，優化最終之共享自行車存量調度與分析系統。綜合上述研究，期望改善共享自行車供需失衡問題，以此提升租賃服務品質、提供營運方進行更好的調度規劃參考。

In order to reduce environmental pollution and minimize resource consumption, the development of Greening Transport Oriented Development (GTOD) has become a globally significant issue. However, in recent years, due to the impact of the COVID-19 pandemic, countries have begun to emphasize social distancing, leading to changes in people's habits of using public transportation. The Public Bicycle System (PBS) is one rapidly developing form of green transportation, offering an effective alternative for both passive and active individual travel. With the widespread adoption of shared bicycles, an issue of supply-demand imbalance has emerged, making the inventory balance of shared bicycles a crucial concern. This study focuses on the urban area of Taipei City and examines the shared bicycle system service provided by its YouBike program. The study conducts an analysis and evaluation of rental patterns and re-balances the inventory of bicycles at rental stations. The research is divided into three parts: rental demand prediction, station inventory risk, and hybrid scheduling recommendations. In the first part, rental demand prediction involves calculating the optimal inventory of bicycles at rental stations. Initially, user rental and return data for different time periods at each rental station are modeled, incorporating the concept of Cyber-Physical-Social Systems (CPSS) and considering social factors such as climate, season, and holidays to analyze user rental demand at each rental station. The XGBoost (eXtreme Gradient Boosting) algorithm is utilized to predict future rental and return rates. The second part focuses on station inventory risk, where the concept of risk levels from the field of management is incorporated. Operational demand indicators are formulated to identify hotspots for bike shortages and station vacancies using numerical indicators. The third part, hybrid scheduling recommendations, utilizes the predicted optimal inventory of bicycles at rental stations and the identified hotspots. Operational strategies are divided into dynamic and static scheduling, optimizing the final shared bicycle inventory scheduling and analysis system. In summary, this comprehensive study aims to address the issue of supply-demand imbalance in shared bicycles, thereby enhancing the quality of rental services and providing operational guidelines for improved scheduling for service providers.