ATD: An Integrated Framework of Network Theory, Data Mining and Partial Differential Equation for Early Detection of Epidemic Outbreaks

尽管医疗技术和疫苗取得了进步，但新出现和再次出现的流行病，如SARS、甲型H1N1流感、禽流感、埃博拉和寨卡病毒，继续构成巨大威胁。及早发现并立即作出反应，对于避免社会影响至关重要。然而，在许多情况下，目前的疫情检测方法和算法无法解释当今可用的丰富社交媒体数据。这些数据为开发改进的监视系统提供了机会。该项目将开发一个新的综合框架，以便根据实时地理标记的推特数据及早发现疫情暴发。这项工作结合了数学和计算机科学领域的科学家的专业知识，并将开发新的算法，以更快地检测(接近实时和局部的)疫情爆发。在局部地区有效地及早发现疫情将大大提高政府机构和卫生组织的认识，促使采取适当的行动来控制和治疗疫情。该项目将极大地提高公众对疫情暴发的卫生意识和防范能力。该项目将开发新的方法和算法，以便利用社交媒体数据及早准确地检测疫情暴发。为此，该项目将在地理标记的Twitter数据的社区检测/聚类和热点话题分析方面引入新的算法。这项工作还将定义有效的距离指标，将基础社交网络的结构、物理距离和旅行信息结合起来，以捕获疫情传播的模式。因此，描述疫情传播时空模式的偏微分方程模型的解被用来提供预测疫情迫在眉睫的早期预警指标。此外，偏微分方程的新算法和新定理将揭示流行病的传播机制。该项目将产生一个新的跨学科框架，包括网络理论、数据挖掘和偏微分方程，用于使用地理标记的Twitter数据进行流行病检测。

Despite advancements in medical technology and vaccines, emerging and reemerging epidemics such as SARS, influenza A (H1N1), avian influenza, Ebola, and Zika continue to pose tremendous threats. Early detection and immediate response are essential to avoid societal repercussions. However, in many cases, current methods and algorithms for epidemic detection cannot account for the wealth of social media data available today. This data provides an opportunity to develop improved surveillance systems. This project will develop a novel integrated framework for early detection of epidemic outbreaks based on real-time geo-tagged Twitter data. The work combines the expertise of scientists in both mathematics and computer science and will develop new algorithms for faster detection (near real-time and localized) of epidemic outbreaks. Effective early detection of epidemics in localized regions will greatly increase governmental agency and health organization awareness, prompting appropriate actions to control and treat epidemics. This project will significantly enhance public health awareness and preparedness against epidemic outbreaks.The project will develop new methodologies and algorithms for early and accurate detection of epidemic outbreaks with social media data. To this end, the project will introduce new algorithms in community detection/clustering and hot topic analysis of geo-tagged Twitter data. The work will also define effective distance metrics that combine the structure of underlying social networks, physical proximity, and travel information to capture the pattern of epidemic spread. As a result, solutions of partial differential equation models, which describe spatio-temporal patterns of epidemic spread, are used to provide an early warning indicator for predicting imminence of an outbreak. In addition, new algorithms and theorems from partial differential equations will reveal epidemic spread mechanisms. The project will produce a new trans-disciplinary framework of network theory, data mining, and partial differential equations for epidemic detection with geo-tagged Twitter data.