Modeling and simulation methods for linked lives in demography

人口统计学中的链接生活模拟模型包括许多代理，以各种角色进行交互，具有详细的决策过程，并动态访问随时间推移而发展的社交网络中的信息。这些模型部分基于已建立的理论、其他模拟模型和数据构建。从这一观察中，建模和仿真的各种挑战确实出现了。开发的，表达建模语言ML3（链接生活建模语言）允许一个简洁的连续时间建模的链接生活模型。 在进一步的案例研究和其他建模语言的背景下，该语言将被评估和改进。有效的执行是彻底开发和分析这些模型的先决条件。因此，应开发并行和近似模拟器。为了有效地将并行模拟应用于链接的生命模型，用于划分社交网络的合适策略是必要的，而近似模拟器在成功地平衡精度和速度方面的有效性依赖于合适的聚合策略。这两种类型的战略都将是研究的主题。为了支持灵活和可复制的仿真实验，ML3和SESSL（通过Scala层的仿真实验规范），一种用于指定和执行仿真实验的语言之间的绑定被开发。分析动态网络的方法应添加到已经提供的分析方法组合中。ML3中的模型和SESSL中的模拟实验的声明性描述应构成自动导出有关促成链接生命模型的各种工件和过程的起源信息的基础。将根据人口统计学和流行病学方面的具体模拟研究对所开发的方法进行评估。

Linked lives simulation models in demography comprise numerous agents, interacting in various roles, with detailed decision processes, and dynamically accessing information in a social network that is evolving over time. These models are built partly based on established theories, other simulation models, and data. From this observation diverse challenges for modeling and simulation do arise. The developed, expressive modeling language ML3 (Modeling Language for Linked Lives) allows a succinct continuous-time modeling of linked lives models. The language shall be evaluated and refined in further case studies and in the context of other modeling languages. An efficient execution is a prerequisite for a thorough development and analysis of such models. Therefore, parallel and approximate simulators shall be developed. For effectively applying parallel simulation to linked lives models, suitable strategies for partitioning the social network are necessary, whereas the effectiveness of approximate simulators in successfully balancing accuracy and speed relies on suitable aggregation strategies. Both types of strategies will be subject of research. To support flexible and replicable simulation experiments, a binding between ML3 and SESSL (Simulation Experiment Specification via a Scala Layer), a language for specifying and executing simulation experiments, was developed. Methods to analyze dynamic networks shall add to the portfolio of analysis methods already provided. The declarative description of models in ML3 and of simulation experiments in SESSL shall form the basis to automatically derive provenance information about the various artifacts and processes that contributed to linked lives models. The developed methods will be evaluated based on concrete simulation studies in demography and epidemiology.