Trajectory Analytics for Data-Driven Predictions and Sequential Decision-Making Under Sequence Uncertainty

序列不确定性下数据驱动预测和序列决策的轨迹分析

• 批准号: RGPIN-2021-04249
• 负责人: Zargoush, Manaf
• 金额: $ 2.16万
• 依托单位: McMaster University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2021
• 资助国家: 加拿大
• 起止时间: 2021-01-01 至 2022-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=742948
• 关键词: Trajectory; Analytics; Data; Driven; Predictions

Background. Healthcare trajectories are a series of healthcare operations processes (such as prevention, diagnostic, and treatment activities) and clinical pathways (such as disease progression) executed in the healthcare system. Identifying, analyzing, and improving the healthcare trajectories has a significant impact on the healthcare costs and patient outcomes by contributing to a) improved system performance, b) controlling disease progression, and c) improved transparency. Healthcare trajectories, which are hidden in the healthcare system's event logs, are highly complex and subject to tremendous variations. Therefore, a data-driven approach, which I refer to as TRajectory Analytics (TRA), is crucial for the characterization, analysis, and improvements of healthcare trajectories through the utilization of the enormous healthcare data and the exponentially-growing computing power we have seen over the last decade. TRA aims at analyzing data and extracting hidden trajectory patterns to generate knowledge regarding the complex sequence of events and actions. Nevertheless, the field of TRA, especially in the healthcare domain, is still in its infancy stages and struggles with various limitations in terms of providing reference data-driven modeling frameworks that encompass multiple analytical scopes, including descriptive, predictive, and prescriptive tasks. Objectives. Addressing the referenced methodological and practical gaps in the context of healthcare TRA is highly aligned with my overarching long-term objectives in developing analytical frameworks for blending predictive and prescriptive analytics to improve data-driven decision-making. In the short-term, I will undertake the following steps as the research objectives over the proposed five-year program of research: 1.Predictive TRA: Developing novel method(s) for discovering and predicting healthcare trajectories from data. 2.Descriptive TRA: Proposing methods for summarizing, visualizing, and reducing the complexity of the identified trajectories in (1) by identifying the clusters of the most common trajectories in optimal fashions. 3.Prescriptive TRA: Proposing optimization frameworks and solution procedures for sequential decision-making based on the explicit utilization of the information gained in (1)-(2). This program is possible through the application of my methodological expertise/experience in healthcare analytics and the availability of rich datasets unique to my lab. Impact. On the scientific side, this research program will deliver several novel methodologies in TRA and illustrates explicit connections between the various scopes of data analytics. On the practical side, it provides the healthcare sector with the tools, scalable to other sectors, that facilitate data-driven decision-making and evidence-based leadership. On the pedagogical side, it provides unique opportunities for training Canadian-trained HQPs in Artificial Intelligence (AI) with expertise in TRA.

背景。医疗保健轨迹是在医疗保健系统中执行的一系列医疗保健操作过程（如预防、诊断和治疗活动）和临床路径（如疾病进展）。识别、分析和改进医疗保健轨迹对医疗保健成本和患者结果有重大影响，有助于a)改进系统性能，b)控制疾病进展，c)提高透明度。隐藏在医疗保健系统事件日志中的医疗保健轨迹非常复杂，并且受到巨大变化的影响。因此，数据驱动的方法，我称之为轨迹分析（TRA），对于通过利用巨大的医疗保健数据和我们在过去十年中看到的指数级增长的计算能力来表征、分析和改进医疗保健轨迹至关重要。TRA旨在分析数据并提取隐藏的轨迹模式，以生成关于复杂事件和动作序列的知识。然而，TRA领域，特别是在医疗保健领域，仍然处于起步阶段，并且在提供包含多个分析范围（包括描述性、预测性和规定性任务）的参考数据驱动建模框架方面受到各种限制。目标。在医疗保健TRA的背景下，解决参考方法和实践方面的差距与我开发分析框架的总体长期目标高度一致，这些分析框架用于混合预测和规范分析，以改进数据驱动的决策。在短期内，我将采取以下步骤作为五年研究计划的研究目标：1。预测性TRA：开发从数据中发现和预测医疗保健轨迹的新方法。2.描述性TRA：通过以最优的方式识别最常见的轨迹簇，提出总结、可视化和降低(1)中已识别轨迹复杂性的方法。3.规定性理论：基于对(1)-(2)中获得的信息的显式利用，提出序列决策的优化框架和求解过程。通过应用我在医疗保健分析方面的方法学专业知识/经验以及我实验室独有的丰富数据集，该计划成为可能。的影响。在科学方面，该研究项目将在TRA中提供一些新颖的方法，并说明各种数据分析范围之间的明确联系。在实践方面，它为医疗保健部门提供了可扩展到其他部门的工具，促进了数据驱动的决策和基于证据的领导。在教学方面，它为加拿大培养的具有TRA专业知识的人工智能（AI） hqp提供了独特的培训机会。