面向流行病防控的大规模人口动态接触网络建模与挖掘方法

Epidemic diseases spread mainly through the contact of population. Given the pathological parameters of pathogens and hosts, the extent of how precisely we can predict the spread trend of diseases using epidemic models will be determined by the extent of how accurately we can construct the contact network of population. Therefore, the study of contact network construction is critical for epidemic disease control and prevention. So far, how to accurately construct the contact network of large-scale population is still an unsolved challenge due to three main reasons. First, the contact network underlying population is hidden and thus it is very difficult to obtain the real-world data that directly and explicitly describe the contact behaviors of large-scale population. Second, the network of human contacts is spatially heterogeneous with respect to their social and demographic scenarios. Third, such network will change temporally at different stages of disease spread, while the existing methods cannot properly characterize and discover the temporal patterns of its dynamics. In this project, we will address the challenge of constructing dynamical contact network over large-scale population from a novel perspective in terms of multiple-source and heterogeneous data mining, by posing and investigating some fundamental research questions that to date have not been addressed in the literature. First, how to represent the spatiotemporal patterns of the dynamical contact network over large-scale meta-population based on tensor model. Second, how to mine the spatial contact patterns of meta-population in different scenarios from multiple-source and heterogeneous data including demographical data, transportation data, wireless sensor data and among others, and thereafter mine the temporal pattern of its dynamical evolution from multiple kinds of epidemic disease surveillance data. On the basis of the above studies, we will further study how to apply the discovered dynamical contact network to address the practical issues of epidemic disease control and prevention. The implementation of this project will deepen, promote and expand the studies and applications of related areas including data mining.

流行病主要通过人与人的接触传播。给定病理学参数后，流行病模型对传播趋势预测的准确程度将取决于人们对人口接触网络的构建精度，因此相关研究对流行病防控具有十分重要的意义。目前，如何准确构建大规模人口的接触网络仍是一个难题，主要因为：接触网络是隐含的，直接描述接触行为的数据很难获取；接触网络是空间异构的，不同场景的接触模式不尽相同；接触网络是动态变化的，现有方法无法刻画和发现其演化模式。针对以上问题，本项目拟从多源异构数据挖掘这一新的角度入手，研究解决大规模人口动态接触网络构建的关键问题：如何基于张量建模大规模复合群体接触网络的时空模式；如何从人口普查、交通、无线传感器等多源异构数据中挖掘出不同场景的复合群体接触模式，进而从多种流行病监控数据中挖掘出复合群体接触模式的动态演化过程。基于以上，进一步研究动态接触网络在流行病防控中的典型应用。本项目的实施将深化拓展数据挖掘等相关领域的研究及应用。

流行病主要通过人与人的接触传播。给定病理学参数后，流行病模型对传播趋势预测的准确程度将取决于人们对人口接触网络的构建精度，因此相关研究对流行病防控具有十分重要的意义。然而，如何准确构建大规模人口的接触网络仍是一个难题。接触网络是隐含的、空间异构的、动态变化的，因此刻画和发现其演化模式面临着重大的挑战。为此，本项目在计算流行病学和复杂网络领域开展了深入的理论和应用研究，本项目的主要创新点是：.1）提出了基于复合群体模型和张量模型的大规模动态接触网络建模方法。该方法能够表示具有空间异构性和多阶段演化特征的大规模复合群体接触结构的时空模式；.2）提出了动态接触网络的空间模式和时间模式的表示和发现方法。.3）面向流行病防控，提出了流行病早期预警、流行病控制策略评估、隐含/缺失流行病监控数据恢复等3种大规模人口动态接触网络的应用模式。.这些工作系统解决了计算流行病学的一些关键问题，深化、拓展了数据挖掘、复杂网络和计算流行病学等方面的理论研究，对流行病的预测与防控具有重要的意义。项目组在IEEE TPAMI、IEEE TKDE、AAAI等人工智能和数据挖掘领域著名期刊和会议上发表署名论文38篇。其中包括：SCI论文24篇，CCF A类论文5篇，CCF B类论文4篇，CCF C类论文10篇，中科院一区论文2篇、二区17篇，出版中文专著2部，译著1部，英文专著章节1章，获得国家专利2项。获得2017年吴文俊自然科学奖二等奖1项。这些工作引起了国内外同行的广泛关注，相关工作多次被国际著名学者，流行病学领域、人工智能领域、数据挖掘领域的顶级会议和期刊论文介绍、评价和对比。

内容获取失败，请点击重试