Easing Parameter Sensitivity Analysis of Netlogo Simulations using SPARTAN

TY - GEN

T1 - Easing Parameter Sensitivity Analysis of Netlogo Simulations using SPARTAN

AU - Alden, Kieran James

AU - Timmis, Jon

AU - Coles, Mark Christopher

PY - 2014

Y1 - 2014

N2 - In attempts to further understand complex systems at an individual level, the application of agent-based modeling is becoming prevalent across a range of academic disciplines. With the advantages of being multi-platform, requiring littleprogramming experience, and supported by a large number of freely available case study examples, Netlogo has become a popular choice as the software tool to apply in the construction of agent-based models. To utilize the constructed model as an informative or predictive tool, statistical analyses can be performed to reveal the influence that a parameter has on simulation behavior, offering an insight into the system under study. Here we demonstrate the integration of Netlogo’s parameter sweep function, Behavior Space, with an extendedversion of SPARTAN, our previously published open source statistical package for performing local and global sensitivity analyses. With the addition of SPARTAN, the researcher can automatically create Netlogo experiment files for bothlocal (individual parameter) and global (latin-hypercube and Fourier frequency) analyses, run these experiments in Netlogo, and receive detailed statistical information on the influence a parameter has on simulation response: vital information for translating a simulation result to a hypothesisgrounded in the system being studied. To ensure our example work is reproducible, we demonstrate use of SPARTAN using the Virus transmission and perpetuation model available in the Netlogo model library.

AB - In attempts to further understand complex systems at an individual level, the application of agent-based modeling is becoming prevalent across a range of academic disciplines. With the advantages of being multi-platform, requiring littleprogramming experience, and supported by a large number of freely available case study examples, Netlogo has become a popular choice as the software tool to apply in the construction of agent-based models. To utilize the constructed model as an informative or predictive tool, statistical analyses can be performed to reveal the influence that a parameter has on simulation behavior, offering an insight into the system under study. Here we demonstrate the integration of Netlogo’s parameter sweep function, Behavior Space, with an extendedversion of SPARTAN, our previously published open source statistical package for performing local and global sensitivity analyses. With the addition of SPARTAN, the researcher can automatically create Netlogo experiment files for bothlocal (individual parameter) and global (latin-hypercube and Fourier frequency) analyses, run these experiments in Netlogo, and receive detailed statistical information on the influence a parameter has on simulation response: vital information for translating a simulation result to a hypothesisgrounded in the system being studied. To ensure our example work is reproducible, we demonstrate use of SPARTAN using the Virus transmission and perpetuation model available in the Netlogo model library.

M3 - Conference contribution

SP - 622

EP - 628

BT - The 14th International Conference on the Synthesis and Simulation of Living Systems

PB - MIT Press

ER -