Streamlining Simulation Experiments with Agent-Based Models in Demography

Abstract

Abstract:In the last decade, the uptake of agent-based modeling in demography and other population sciences has been slowly increasing. Still, in such areas, where traditional data-driven, statistical approaches prevail, the hypothesis-driven design of agent-based models leads to questioning the validity of these models. Consequently, suitable means to increase the confidence into models and simulation results are required. To that end, explicit, replicable simulation experiments play a central role in model design and validation. However, the analysis of more complex models implies executing various experiments, each of which combines various methods. To streamline these experimentation processes a flexible computational simulation environment is necessary. With a new binding between SESSL -an internal domain-specific language for simulation experiments -and ML -a simulator for linked lives designed specifically for agent-based demographic models -we cater for these objectives and provide a powerful simulation tool. The proposed approach can serve as a foundation for current e orts of employing advanced and statistical model analysis of agent-based demographic models, as part of a wider process of iterative model building. We demonstrate its potential in specifying and executing di erent experiments with a simple model of return migration and a more complex model of social care.

Developing Agent-Based Migration Models in Pairs

“…ML3 models are interpreted by a Java-based simulator (Reinhardt and Uhrmacher 2017) using a variation of Gillespie's Stochastic Simulation Algorithm (Gillespie 1976). The language has been successfully applied to models of human migration and decision processes (Warnke et al 2017;Reinhardt et al 2018). However, the collection and adaptation of knowledge, a central part of this model, was not part of previous models.…”

“…In ML3, simulation experimentation is realized via a binding to SESSL (Ewald and Uhrmacher 2014), a domain specific language for specification of simulation experiments embedded into the GPL Scala. SESSL, allows to specify simulation experiments in a declarative and executive manner, including, e.g., parallel execution of replications, parameterization and experimental design, some methods for statistical analysis, and export of simulation results for analysis with external tools (Reinhardt et al 2018). Experiments specified in SESSL are directly executable Scala code.…”

Using Multiple Scale Space-Time Patterns to Determine the Number of Replicates and Burn-In Periods in Spatially Explicit Agent-Based Modeling of Vector-Borne Disease Transmission

“…Therefore, multiple simulation runs need to be performed to have proper sample sizes for hypothesis testing (e.g., what-if scenario analysis) and validating simulation models [14][15][16]. The number of simulation runs varies according to the researchers, and it often ranges from ten to 100 [17][18][19][20][21][22]. The question about how many simulation runs are required still needs to be answered.…”

Application of Agent-based approach to the Modeling of Processes in the Field of Migration Policy

“…Данная статья посвящена анализу существующего опыта применения агенториентированного подхода к моделированию миграционных процессов, проводимому с целью дальнейшей разработки комплекса агентных моделей, имитирующих процессы в области миграционной политики. С помощью технологий SESSL [6] и ML3 [17] в работе [14] была разработана модель, имитирующая процессы возвратной миграции. Мигранты обладают рядом характеристик, влияющих на их финансовое состояние, а также желание возвращения на родину.…”