"Agent based modeling of urban-level, contact-based infectious disease spread"

2009/2010年全球H1N1流感大流行使研究重新关注感染传播和控制的许多方面，包括疾病的流行病学、疫苗和其他药物措施的开发和部署，以及公共卫生和应急管理措施。了解和预测大流行性流感如何在人群中传播是至关重要的，因为它可能对公共卫生和一个区域的整体经济造成毁灭性后果。本研究计划的目标是开发一个精确、准确和可扩展的基于主体的建模（ABM）框架，通过该框架来模拟感染传播的复杂社会动态，并模拟城市或地区规模的公共卫生干预措施。这项工作超越了之前的努力，专注于将真实数据集成到人们（代理人）的运动、行为和相互作用的特征概况中，以达到高度的特异性和保真度。在这个项目中，真实数据包括合并现有的真实数据来定义模型地形和代理配置文件，开发新的数据工具来建模细粒度和中粒度的社交网络，以及集成实时数据馈送。此外，通过整合机器学习模型或由模型自动生成的代理策略，以最大限度地减少感染传播，这项工作超越了以前的努力。该模型将用于模拟感染控制措施对城市和较大地区流感样疾病的影响。这项工作可作为公共卫生和政策分析人员以及应急管理人员的决策支持工具，立即得到应用。这项工作的独特之处在于，ABM建模和仿真方法将社会系统的现实世界复杂性纳入了一定程度的敏感性和保真度，这是其他方法（如数学建模）在理解复杂现象方面无法比拟的。ABM的优势和新颖性通过整合高性能计算的潜力，研究人员可以获得的真实社会数据的数量不断增加，以及机器学习的包含而进一步放大。

The worldwide H1N1 influenza pandemic in 2009/2010 renewed research attention to the many aspects of infection spread and control, ranging from the epidemiology of the illness, development and deployment of vaccines and other pharmaceutical measures, and public health and emergency management measures. It is critical to understand and predict how pandemic influenza spreads within a population, due to the devastating potential consequences to public health and to the overall economy of a region. This research program's objective is to develop a precise, accurate, and scalable Agent-Based Modeling (ABM) framework by which to simulate the complex social dynamics of infection spread and to simulate public health interventions at the scale of a city or region. The work goes beyond previous efforts by focusing on the integration of real data to model characteristic profiles of people's (agents') movements, behaviours, and interaction with one another to a high level of specificity and fidelity. In this program, real data includes the incorporation of existing real data to define model topographies and agent profiles, the development of new data tools to model fine- and medium-grained social networks, and the integration of real-time data feeds. Additionally, the work goes beyond previous efforts by integrating models of machine learning, or agent policy that will be automatically generated by the model to minimize infection spread. The model will be used to simulate the impacts of infection control measures for to influenza-like illness in cities and larger regions. The work has immediate applications as a decision support tool to public health and policy analysts and to emergency management. The work is unique in that the ABM approach to modeling and simulation incorporates the real-world complexity of social systems to a degree of sensitivity and fidelity that other approaches such as mathematical modeling cannot match in terms of understanding a complex phenomenon. The advantages and novelty of ABM are further amplified by integrating the potential of high performance computing, the increasing amounts of real social data becoming available to researchers, and the inclusion of machine learning.