Variance in System Dynamics and Agent Based Modelling Using the SIR Model of Infectious Disease

TL;DR

This paper compares System Dynamics and Agent Based Modelling for the SIR infectious disease model, showing that Agent Based Modelling better captures natural variability in disease spread simulations.

Contribution

It demonstrates that Agent Based Modelling inherently captures variability in epidemiological simulations better than traditional System Dynamics with Monte-Carlo methods.

Findings

Agent Based Modelling shows greater variability in results.

System Dynamics with Monte-Carlo underestimates natural variation.

Agent Based approach aligns more closely with empirical data.

Abstract

Classical deterministic simulations of epidemiological processes, such as those based on System Dynamics, produce a single result based on a fixed set of input parameters with no variance between simulations. Input parameters are subsequently modified on these simulations using Monte-Carlo methods, to understand how changes in the input parameters affect the spread of results for the simulation. Agent Based simulations are able to produce different output results on each run based on knowledge of the local interactions of the underlying agents and without making any changes to the input parameters. In this paper we compare the influence and effect of variation within these two distinct simulation paradigms and show that the Agent Based simulation of the epidemiological SIR (Susceptible, Infectious, and Recovered) model is more effective at capturing the natural variation within SIR compared to an equivalent model using System Dynamics with Monte-Carlo simulation. To demonstrate this effect, the SIR model is implemented using both System Dynamics (with Monte-Carlo simulation) and Agent Based Modelling based on previously published empirical data.