Predictive and Prescriptive Analytics for Equitable and Efficient Service Provision in Smart Cities

以公平和有效的方式为居民提供服务是加拿大和世界各地政策制定者的首要任务。在私营部门，借助零工的力量，在线平台发明了服务客户的新方式，利用了数据分析的力量。然而，如果没有适当的算法设计，盲目使用数据可能会导致次优和有偏差的结果。该提案旨在开发基于预测和规范分析工具的数据驱动数学模型和算法，以支持服务运营管理，确保所有公民的公平和效率。具体而言，该提案将重点关注三个密切相关的研究课题，这些课题在文献中尚未得到很好的解决：(1)考虑公平因素的数据驱动的设施选址优化；(2)众包资源配置，保证应急物资的准时送达；(3)不确定条件下的联合容量管理与调度优化。在这三个项目中，我们将探索可以从数据中驱动决策的新模型和方法。预测和规范分析将仔细集成，以实现具有样本外性能保证的理论上最佳解决方案。为了考虑实际场景中需求和供应的时空不确定性，我们还将研究单期和多期设置下的稳健优化技术。由于新的问题结构可能会随着权益约束的增加而出现，我们将设计易于实现和可扩展的算法来有效地解决由此产生的数学程序。开发的方法和算法将通过从公共资源和行业合作伙伴收集的数据，在公共服务、医疗保健和最后一英里物流等实际问题上进行测试。基于我们丰富的合作经验，我们将与不同的利益相关者（市政府、在线平台和非营利组织）寻求实施机会。我们期待研究成果能为提高加拿大居民的生活质量带来理论突破和有见地的政策建议。

Providing services to residents in an equitable and efficient manner is a top priority for policymakers in Canada and around the world. In the private sector, online platforms empowered by gig workers have invented new ways of serving customers, harnessing the power of data analytics. However, without an adequate design of algorithms, a blind use of data may lead to suboptimal and biased outcomes. This proposal aims to develop data-driven mathematical models and algorithms built on predictive and prescriptive analytics tools to support service operations management, ensuring equity and efficiency for all citizens. Specifically, the proposal will focus on three closely related research topics that have not been well addressed in the literature: (1) data-driven facility location optimization with equity considerations; (2) crowd-sourcing resource allocation for on-time delivery of emergency supplies; and (3) joint capacity management and dispatching optimization under uncertainty. In the three projects, we will explore new models and methodologies that can drive decisions from data. Predictive and prescriptive analytics will be integrated carefully to enable theoretically optimal solutions with out-of-sample performance guarantees. To account for the spatiotemporal uncertainty of demand and supply in practical scenarios, we will also investigate robust optimization techniques in both single-period and multi-period settings. Because new problem structures can arise with the addition of equity constraints, we will design easily implementable and scalable algorithms to solve the resulting mathematical programs efficiently. The developed methods and algorithms will be tested on real-world problems in public services, healthcare, and last-mile logistics, with data collected from public sources and industry partners. Based on our extensive collaboration experience, we will seek implementation opportunities with different stakeholders (municipal governments, online platforms, and non-profit organizations). We expect the research outcome will bring theoretical breakthroughs and insightful policy suggestions to improve the life quality of Canadian residents.