The high demand for radiotherapy services, combined with the limited capacity of available resources, patient unpunctuality, and series of appointments, makes Patient Appointment Scheduling (PAS) in radiotherapy centers very challenging. Although most centers use a First-Come-First-Serve (FCFS) policy for appointment scheduling, this approach does not consider patients’ behaviors, and consequently, it performs poorly. This type of inappropriate scheduling usually leads to inefficiency at the center and/or patient dissatisfaction. This study provides a data-driven approach to patient appointment scheduling to deal with the challenges mentioned above, and it considers patients’ histories of unpunctuality, including the amount of time they are usually late and whether they will miss the appointment. This study first employs data-mining techniques to predict patients’ behaviors and then incorporates them into PAS. In addition, it presents a novel double-stage prioritization method that considers both patients’ gradual health improvement during the treatment process and any treatment prolongation that occurs. These predictions and priorities are then utilized in the developed Mixed Integer Linear Programming (MILP) model to determine the optimal sequence of patients for treatment. The developed model also considers no-show patients and rearranges their makeup session(s) to meet their service requirements. Lastly, the proposed approach is applied to two business configurations (i.e., single-server and multi-server radiotherapy centers) to highlight its advantages and demonstrate its performance against the current policy. The results reveal that employing the developed model improves the center’s total cost by up to 30%.

对放射治疗服务的高需求，加上可用资源的有限能力、患者不守时和一系列预约，使得放射治疗中心的患者预约安排 (PAS) 非常具有挑战性。尽管大多数中心使用先到先服务 (FCFS) 策略进行预约安排，但这种方法没有考虑患者的行为，因此效果不佳。这种不恰当的安排通常会导致中心效率低下和/或患者不满意。本研究为患者预约安排提供了一种数据驱动的方法来应对上述挑战，并考虑了患者不守时的历史，包括他们通常迟到的时间量以及他们是否会错过预约。本研究首先采用数据挖掘技术来预测患者的行为，然后将其纳入 PAS。此外，它提出了一种新颖的双阶段优先排序方法，该方法既考虑了患者在治疗过程中的逐渐健康改善，也考虑了发生的任何治疗延长。然后在开发的混合整数线性规划 (MILP) 模型中使用这些预测和优先级来确定患者的最佳治疗顺序。开发的模型还考虑了未出现的患者并重新安排他们的化妆会议以满足他们的服务要求。最后，将所提出的方法应用于两种业务配置（即单服务器和多服务器放射治疗中心），以突出其优势并展示其针对当前政策的性能。