CAREER: Improving Operational Decision Making with Predictive Information and Data

这项教师早期职业发展（Career）资助通过研究提高医疗保健和其他关键服务资源规划和运营效率的方法来促进国民健康和福利。该奖项扩展了我们对数据驱动方法的理解，这些方法可以提供更好的患者流和人员配置模型。随着数据的可用性和计算能力的不断增强，人们对数据支持的决策的认识和欣赏也越来越高。然而，在确定哪种类型的预测信息最有价值以及随后对系统性能进行量化方面仍然存在挑战。研究结果将提供结构性见解和有效的政策，以提高系统效率、消费者体验和护理质量。教育任务的目的是让学生接触到提高服务和医疗保健行业的运营效率和降低风险的挑战。随附的教育计划有助于培养工程师和研究人员，他们将在数据知情的运筹学方面具备扎实的理论背景和实践见解。通过教育活动，首席研究员致力于促进未被充分代表的群体参与工程。这项研究包含两条主线。第一部分研究了如何利用预测信息来提高系统性能。目标是建立一个理论框架来评估不同类型信息的有效性，并量化信息准确性的影响。从随机过程的渐近分析技术将被扩展到纳入和分析预测信息的影响。此外，还将构建先进的数据知情策略，这些策略可以显著提高性能并对预测误差具有鲁棒性。重点是资源有限的环境，需要仔细模拟和研究外部性。在医疗保健应用的推动下，面对多尺度不确定性的瞬态性能分析和优化的新范式将被开发出来。第二个线程研究如何利用数据来建立更好的随机模型。目标是开发一个数据驱动的建模框架，提供系统操作中显著权衡的准确量化。将计量经济学与随机建模相结合，解决模型校准、因果推理和绩效评估方面的关键挑战。基于模型的高保真特性，将提供易于实现的说明性解决方案。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This Faculty Early Career Development (CAREER) grant promotes the national health and welfare by studying methods to improve the efficiency of resource planning and operations in healthcare and other critical services. This award extends our understanding of data-driven methods that can provide better models of patient flow and staffing. With the growing availability of data and computational powers, there is an increasing awareness and appreciation of data-backed decision making. However, challenges remain in identifying what types of predictive information are most valuable, and subsequently in quantifying system performance. The findings will provide structural insights and effective policies to improve system efficiency, consumer experience, and quality of care. The educational mission aims to expose students to the challenges of improving operational efficiency and reducing risk in service and healthcare industries. The accompanying educational plan facilitates training of engineers and researchers who will be equipped with both solid theoretical backgrounds as well as practical insights in data-informed operations research. Through the educational activities, the Principal Investigator is committed to promoting the participation of underrepresented groups in engineering. This research contains two main threads. The first studies the use predictive information to achieve better system performance. The objective is to develop a theoretical framework to evaluate the effectiveness of different kinds of information and to quantify the impact of the information’s accuracy. Techniques from asymptotic analysis of stochastic processes will be extended to incorporate and analyze the impact of predictive information. Moreover, advanced data-informed policies that lead to substantial performance improvement and are robust to prediction errors will be constructed. The focus is on limited-resource environments where externalities need to be carefully modeled and studied. Motivated by healthcare applications, new paradigms for transient performance analysis and optimization in face of multiple scales of uncertainty will be developed. The second thread studies how to utilize data to build better stochastic models. The objective is to develop a data-driven modeling framework that provides accurate quantifications of salient tradeoffs in system operations. Key challenges in model calibration, causal inference, and performance evaluation will be addressed by combining tools from econometrics with stochastic modeling. Prescriptive solutions that can be readily implemented will be provided based on the high-fidelity nature of the models.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.