Algorithms for Pathway Operations: Theory and Application to Museums

随着市场和智能供应链的发展，对数据驱动的运营进行了前所未有的研究。资源分配和个性化算法使我们能够在零售、运输和医疗保健等行业更有效地匹配供需。然而，这些算法通常针对单个用户平台交互进行优化，例如，在给定产品类别中进行购买决策。平台运营的一个主要问题是开发用户体验的整体观点，并由跟踪纵向行为的新数据源（如移动跟踪数据和点击流数据）提供支持。因此，我建议设计算法来优化多点触控操作平台中重复用户交互的参与度。作为案例研究，我将探讨“路径操作”的一个高度相关但未知的应用：博物馆和画廊等艺术机构的物理和数字布局的游客体验。我的研究工作将从算法设计的理论进步到与博物馆合作的应用工作。首先，我将为路径选择模型开发一个新的预测效用最大化框架，它可以从基于轨迹的路径数据中进行估计。其次，我将研究以路径选择模型为核心的算法的设计和严格分析，包括分类的新公式、在线匹配和多臂强盗问题。第三，基于与一家大型博物馆的持续合作，我将与实证领域的研究人员一起转向这些算法的模拟和实现。我们将生成开源代码，帮助博物馆等机构为游客创造新的体验，从推荐的物理路线到个性化的虚拟展览。这项研究工作将有助于更广泛的倡议，为我们的艺术遗产建立一个互动、非侵入性和以博物馆为中心的数字生态系统。

With the growth of marketplaces and smart supply chains, there has been unprecedented research on data-driven operations. Algorithms for resource allocation and personalisation allow us to match supply with demand more efficiently in industries such as retail, transportation, and healthcare. Yet, these algorithms generally optimise for a single user-platform interaction, e.g., a purchase decision in a given product category. A major issue for platform operations is to develop a holistic viewpoint on user experience, supported by new data sources tracking longitudinal behaviour such as mobile trace data and clickstream data. Thus, I propose to design algorithms that optimise engagement for repeated user interactions in multi-touch operational platforms. As a case study, I will explore a highly relevant, yet uncharted application of "pathway operations": visitor experience in art institutions such as museums and galleries across physical and digital layouts. My research work will range from theoretical advancements in algorithm design to applied work in collaboration with museums. First, I will develop a new predictive utility maximisation framework for pathway choice models, which can be estimated from trajectory-based pathway data. Second, I will study the design and rigorous analysis of algorithms incorporating pathway choice models at their core, including novel formulations of assortment, online matching and multi-armed bandit problems. Third, building on an ongoing collaboration with a large museum, I will turn to simulations and implementations of these algorithms together with researchers in empirical fields. We will generate open-source code to help institutions like museums craft new experiences for their visitors, ranging from recommended physical routes to personalised virtual exhibitions. This research work will contribute to broader initiatives for an interactive, non-invasive and museum-centric digital ecosystem for our artistic heritage.