Exploring the behavior space of agent-based simulation models using random forest metamodels and sequential sampling

Edali, Mert; Yücel, Gönenç

doi:10.1016/j.simpat.2018.12.006

Cited by 26 publications

(24 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this paper, we aim to demonstrate the efficacy of the proposed workflow in analysing an empirically grounded and relatively large‐scale IBM. This paper also extends our previous work (Edali & Yücel, 2019), which deals with metamodel training through adaptive sampling, by proposing an additional rule extraction step from RF metamodels to enhance understanding into input–output relations. We will be focusing on the FluTE model (Chao, Halloran, Obenchain, & Longini, 2010), which is a large‐scale individual‐based influenza epidemic model, and using the proposed approach on this model in order to understand the pandemic dynamics and develop insights about the impact of policy parameters on the results.…”

Section: Introductionmentioning

confidence: 54%

“…The machine learning literature offers a diverse set of tools and algorithms, which are capable of overcoming the limitations of both manual and DoE‐based analysis techniques, to perform model analysis in a systematic way. Among these tools, supervised learning algorithms aim to build a mapping from simulation model inputs to the outputs by learning a linear/nonlinear functional relationship from the input–output data obtained from the simulation model results (Alpaydin, 2014; Edali & Yücel, 2018; Edali & Yücel, 2019). The function inferred from the simulation input–output data can be considered as an estimated representation of the input–output relationship of the original SDM or IBM and is also called a metamodel, a surrogate model, a proxy, an emulator or a response surface (Kleijnen, 2009; Kleijnen, Sanchez, Lucas, & Cioppa, 2005; Kleijnen & Sargent, 2000).…”

Section: Introductionmentioning

confidence: 99%

“…The function inferred from the simulation input–output data can be considered as an estimated representation of the input–output relationship of the original SDM or IBM and is also called a metamodel, a surrogate model, a proxy, an emulator or a response surface (Kleijnen, 2009; Kleijnen, Sanchez, Lucas, & Cioppa, 2005; Kleijnen & Sargent, 2000). In an attempt towards proposing a systematic and, perhaps, automated model analysis approach for IBMs, several former studies utilize metamodels of different sorts, which guide the process of behaviour space exploration and/or of relationship identification between model inputs and outputs (Chérel, Cottineau, & Reuillon, 2015; Edali & Yücel, 2018; Edali & Yücel, 2019; Edmonds, Little, Lessard‐Phillips, & Fieldhouse, 2014; Lamperti, Roventini, & Sani, 2018).…”

Section: Introductionmentioning

confidence: 99%

“…Then, we employed DTs to present the knowledge embedded in the black‐box SVR metamodel in the form of intelligible IF–THEN rules. In subsequent studies, we also showed that the parameter sampling process can be driven by the information derived from random forest (RF) metamodels, which results in an adaptive parameter space exploration focusing on the ‘interesting parts’ of the behaviour space of an IBM (Edali & Yücel, 2019). As a result, the aforementioned pieces fit together to form a formal model analysis process based on RF metamodels, where the model is trained through an adaptive and iterative sampling process, and the information captured by the resulting RF is summarized in the form of rules that describe the behavioural regularities of the model being analysed.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Analysis of an individual‐based influenza epidemic model using random forest metamodels and adaptive sequential sampling

Edali

Yücel

2020

Syst Res Behav Sci

Self Cite

View full text Add to dashboard Cite

This study proposes a three‐step procedure for the analysis of input–response relationships of dynamic models, which enables the analyst to develop a better understanding about the dynamics of the system. The main building block of the procedure is a random forest metamodel capturing the input–output relationships. We utilize an active learning approach as the second step to improve the accuracy of the metamodel. In the last step, we develop a novel way to present the information captured by the metamodel as a set of intelligible IF–THEN rules. For illustration, we use the FluTE model, which is an individual‐based influenza epidemic model. We observe that the number of daily applicable vaccines determines the success of an intervention strategy the most. Another critical observation is that when the daily available vaccines are constrained, nonpharmaceutical strategies should be incorporated to reduce the extent of the outbreak.

show abstract

Section: Introductionmentioning

confidence: 54%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Analysis of an individual‐based influenza epidemic model using random forest metamodels and adaptive sequential sampling

Edali

Yücel

2020

Syst Res Behav Sci

Self Cite

View full text Add to dashboard Cite

show abstract

“…RandomForest (RF) is an efficient procedure that is robust to overfitting and flexible regarding covariate interactions and nonlinearity [50]. Furthermore, RF affords easy and reliable methods for interpretation, which has led to its recent exploitation in individual-based model metamodelling [52].…”

Section: Regression Analysesmentioning

confidence: 99%

The boon and bane of boldness: movement syndrome as saviour and sink for population genetic diversity

et al. 2020

View full text Add to dashboard Cite

Background: Many felid species are of high conservation concern, and with increasing human disturbance the situation is worsening. Small isolated populations are at risk of genetic impoverishment decreasing within-species biodiversity. Movement is known to be a key behavioural trait that shapes both demographic and genetic dynamics and affects population survival. However, we have limited knowledge on how different manifestations of movement behaviour translate to population processes. In this study, we aimed to 1) understand the potential effects of movement behaviour on the genetic diversity of small felid populations in heterogeneous landscapes, while 2) presenting a simulation tool that can help inform conservation practitioners following, or considering, population management actions targeting the risk of genetic impoverishment. Methods: We developed a spatially explicit individual-based population model including neutral genetic markers for felids and applied this to the example of Eurasian lynx. Using a neutral landscape approach, we simulated reintroductions into a three-patch system, comprising two breeding patches separated by a larger patch of differing landscape heterogeneity, and tested for the effects of various behavioural movement syndromes and founder population sizes. We explored a range of movement syndromes by simulating populations with various movement model parametrisations that range from 'shy' to 'bold' movement behaviour. Results: We find that movement syndromes can lead to a higher loss of genetic diversity and an increase in between population genetic structure for both "bold" and "shy" movement behaviours, depending on landscape conditions, with larger decreases in genetic diversity and larger increases in genetic differentiation associated with bold movement syndromes, where the first colonisers quickly reproduce and subsequently dominate the gene pool. In addition, we underline the fact that a larger founder population can offset the genetic losses associated with subpopulation isolation and gene pool dominance.(Continued on next page)

show abstract

Quantifying the Safe Operating Space for Land-System SDG Achievement via Machine Learning Meta-Modelling and Scenario Discovery

Khan

Nazari

Thiruvady

et al. 2022

Preprint

View full text Add to dashboard Cite

We developed a machine learning based meta-model to identify sustainability pathways through rapid scenario generation and defined the safe operating space for achieving them via scenario discovery. We trained a meta-model to replicate the Land-Use Trade-Offs integrated model of the Australian land system. Latin hypercube sampling was used to create many scenarios exploring the impact of uncertainties in key drivers including future socio-economic development, climate change mitigation, and agricultural productivity at a granular level. Economic and environmental impacts were evaluated against nationally downscaled SDG targets. Scenario discovery revealed new pathways to achieving five SDG targets for 2050 which required crop yield increases above 1.78 times, a carbon price above 100 AU$ tCO2-1, a > 9% biodiversity levy on carbon plantings, and carefully regulated land-use policy. Machine learning based meta-modelling teamed with scenario discovery revealed the policy and scenario settings required for a sustainable future for the Australian land sector.

show abstract

Exploring the behavior space of agent-based simulation models using random forest metamodels and sequential sampling

Cited by 26 publications

References 57 publications

Analysis of an individual‐based influenza epidemic model using random forest metamodels and adaptive sequential sampling

Analysis of an individual‐based influenza epidemic model using random forest metamodels and adaptive sequential sampling

The boon and bane of boldness: movement syndrome as saviour and sink for population genetic diversity

Quantifying the Safe Operating Space for Land-System SDG Achievement via Machine Learning Meta-Modelling and Scenario Discovery

Contact Info

Product

Resources

About