ATD: Algorithms for Real-time Dynamic Risk Identification with Statistical Confidence

数字技术的最新进展，如宽带网络，在线市场，大型供应链和物流网络，广泛的智能手机使用，可穿戴设备和数字健康技术，促进了近实时，高分辨率数据集的生成和存储。这些数据集从大量主题中以高频率连续可用，跨越各个领域，包括医疗保健，医学，移动的健康，供应链和网络监控。这种类型的数据收集，通常被称为流数据，带来了技术的范式转变，并通过监控动态数据和及时做出连续决策，为实时威胁检测提供了重要机会。该研究项目通过开发用于实时动态风险识别的算法来充分利用流数据研究的潜力，该算法充分探索了海量数据流的独特特征。该项目将为研究生提供研究培训机会。研究人员旨在开发流数据中实时风险检测的算法和统计方法，特别是在电子医疗记录，移动的健康和供应链领域。所开发的方法允许动态修改统计模型，有效存储汇总统计数据，并在新数据到达时准确检测威胁和异常行为。通过合并相关和非相同分布的样本，该项目摆脱了简单化的模型，并采用了新的框架来反映领域问题和数据现实。这些方法将用于解决艾滋病毒预防、移动的健康与抑郁症以及供应链中断等方面的重大科学问题。该项目将为动态风险识别创建一个统一的框架，该框架可以很容易地纳入各个学科，促进与主题科学家的合作，并通过教育计划让学生参与最先进的研究。该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估来支持。

Recent advancements in digital technologies, such as wide-bandwidth networks, online marketplaces, large supply chain and logistics networks, widespread smartphone usage, wearable devices, and digital health technologies, have facilitated the generation and storage of near-real-time, high-resolution datasets. These datasets are sequentially available at a high frequency from a large number of subjects, spanning various fields including healthcare, medicine, mobile health, supply chain, and network monitoring. This type of data collection, commonly referred to as streaming data, has brought about a paradigm shift in technology and presents significant opportunities for real-time threat detection by monitoring data in motion and making continuous decisions in a timely manner. This research project leverages the potential of streaming data research by developing algorithms for real-time dynamic risk identification that fully explore the unique features of massive data streams. The project will provide research training opportunities for graduate students.The investigators aim to develop algorithms and statistical methods for real-time risk detection in streaming data, particularly in the domains of electronic medical records, mobile health, and supply chain. The developed approaches allow for dynamic revision of statistical models, efficient storage of summary statistics, and accurate detection of threats and abnormal behaviors as new data arrives. By incorporating dependent and non-identically distributed samples, the project moves away from simplistic models and embraces new frameworks to reflect domain problems and data realities. The approaches will be applied to address significant scientific questions in HIV prevention, mobile health with depression disorders, and supply chain disruptions. The project will create a unified framework for dynamic risk identification that can be readily incorporated into various disciplines, fostering collaborations with subject-matter scientists, and involving students in state-of-the-art research through educational initiatives.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.