We study the delivery of mobile medical services and in particular, the optimization of the joint stop location selection and routing of the mobile vehicles over a repetitive schedule consisting of multiple days. Considering the problem from the perspective of a mobile service provider company, we aim to provide the most revenue to the company by bringing the services closer to potential customers. Each customer location is associated with a score, which can be fully or partially covered based on the proximity of the mobile facility during the planning horizon. The problem is a variant of the team orienteering problem with prizes coming from covered scores. In addition to maximizing total covered score, a secondary criterion involves minimizing total travel distance/cost. We propose a data-driven optimization approach for this problem in which data analyses feed a mathematical programming model. We utilize a year-long transaction data originating from the customer banking activities of a major bank in Turkey. We analyze this dataset to first determine the potential service and customer locations in Istanbul by an unsupervised learning approach. We assign a score to each representative potential customer location based on the distances that the residents have taken for their past medical expenses. We set the coverage parameters by a spatial analysis. We formulate a mixed integer linear programming model and solve it to near-optimality using Cplex. We quantify the trade-off between capacity and service level. We also compare the results of several models differing in their coverage parameters to demonstrate the flexibility of our model and show the impact of accounting for full and partial coverage.

中文翻译：

---

用于路由移动医疗设施的数据驱动优化框架

我们研究了移动医疗服务的提供，特别是在由多天组成的重复计划中优化联合停靠位置选择和移动车辆的路线。从移动服务提供商的角度考虑这个问题，我们的目标是通过使服务更接近潜在客户来为公司提供最大的收入。每个客户位置都与一个分数相关联，在规划范围内，可以根据移动设施的接近程度来完全或部分覆盖该分数。这个问题是团队定向运动问题的一个变体，奖品来自覆盖的分数。除了最大限度地提高总覆盖分数外，次要标准还涉及最大限度地减少总旅行距离/成本。我们针对这个问题提出了一种数据驱动的优化方法，其中数据分析提供了一个数学规划模型。我们利用来自土耳其一家主要银行的客户银行活动的长达一年的交易数据。我们分析此数据集，首先通过无监督学习方法确定伊斯坦布尔的潜在服务和客户位置。我们根据居民过去医疗费用的距离为每个具有代表性的潜在客户位置分配一个分数。我们通过空间分析设置覆盖参数。我们制定了一个混合整数线性规划模型，并使用 Cplex 将其求解为接近最优。我们量化了容量和服务水平之间的权衡。