面向动态接触网络的传染病早期发现方法研究

发布时间：2018-05-22 20:16

本文选题：计算流行病学 + 复杂网络　；参考：《吉林大学》2017年硕士论文

【摘要】：传染病早期发现是计算流行病学和复杂网络科学的研究热点。传染病爆发具有不可预知、传播速度快、感染范围广和难以有效控制等特点,每次发生都会给人类社会造成巨大的生命和财产损失。设计有效的传染病早期发现方法是传染病防控的有效手段。尽早的预测出传染病爆发的趋势可为相关部门提供充分的应对时间,提前采取防控措施,将其危害降至最低。现有的传染病早期发现研究大都面向静态接触网络,基于静态网络的拓扑属性设计早期发现策略。然而,人与人之间的接触往往是动态变化的,静态接触网络不能很好的刻画现实世界中的实际接触行为。静态网络模型与真实接触模式的偏差会降低早期发现的准确性。在此背景下,本文开展了面向动态接触网络的传染病传播过程建模和早期发现方法研究,提出3个适用于动态网络的早期发现方法,具体完成如下两个主要工作。1)借鉴面向动态网络的传染病免疫策略,提出了2种针对动态接触网络的传染病早期发现方法,这2种方法都是根据网络的时序性特点选择需要重点监控的监控目标。基于真实数据集对它们的性能进行了实证研究和定量分析。实验结果表明:借鉴动态免疫策略提出的早期发现方法可以实现传染病的早期发现工作,并且得到的预测结果较优于现有的静态网络下最有效的早期发现方法。2)以上2种策略在进行早期发现时需要重复处理接触数据,计算开销大。针对该问题,本文改善了以上两个方法的不足,进一步提出了简化数据处理的传染病早期发现方法。不需要随机选择部分个体再进行数据统计,节省了时间开销,提高了传染病的早期发现能力。该方法可根据易于获得的数据(局部接触网络结构和高活度个体间的接触时序信息)有效预测传染病的爆发时间。基于真实数据集的实证研究表明:该方法得到的预测结果优于上述2种基于动态网络的早期发现方法。本文工作是国际上第一个面向动态接触网络的传染病早期发现方法研究。该工作进一步完善了现有的面向网络的传染病早期发现方法研究的局限性,为后续该方向的研究起到了更好的推进作用。
[Abstract]:The early discovery of infectious diseases is a hot topic in computational epidemiology and complex network science. The outbreak of infectious diseases has the characteristics of unpredictable, rapid transmission, wide range of infection and difficult to effectively control, each occurrence will cause huge loss of life and property to human society. To design effective methods for early detection of infectious diseases is an effective means to prevent and control infectious diseases. Early prediction of the trend of infectious disease outbreak can provide adequate response time for relevant departments, take preventive and control measures ahead of time, and reduce its harm to the minimum. Most of the existing researches on early detection of infectious diseases are oriented to static contact networks, and the topology properties of static networks are used to design early detection strategies. However, the contact between people is often dynamic, static contact network can not well describe the actual contact behavior in the real world. The deviation between static network model and real contact mode will reduce the accuracy of early detection. In this context, the modeling and early detection methods of infectious disease transmission process oriented to dynamic contact network are studied in this paper, and three early detection methods suitable for dynamic network are proposed. The following two main works have been accomplished: 1) two methods for early detection of infectious diseases based on dynamic contact network are proposed, which draw lessons from the immunization strategy of infectious diseases oriented to dynamic network. These two methods are based on the temporal characteristics of the network to select monitoring targets that need to be monitored. Based on the real data set, the performance of them is analyzed quantitatively and empirically. The experimental results show that the early detection of infectious diseases can be realized by using the early detection method proposed by the dynamic immune strategy. And the predicted results are better than the most effective early detection methods in static network. 2) the above two strategies need to deal with the contact data repeatedly in the process of early detection, and the computation cost is high. In order to solve this problem, this paper improves the shortcomings of the above two methods, and further proposes an early detection method for infectious diseases which simplifies data processing. No random selection of individuals is needed for data statistics, which saves time and improves the ability of early detection of infectious diseases. This method can effectively predict the outbreak time of infectious diseases based on the easily available data (local contact network structure and contact time series information between individuals with high activity). An empirical study based on real data sets shows that the prediction results obtained by this method are superior to those of the two dynamic network-based early discovery methods. This paper is the first international research on the early detection of infectious diseases for dynamic contact networks. This work further improves the limitations of the existing network oriented early detection methods of infectious diseases, and plays a better role in further research in this direction.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R181;O157.5

【相似文献】