Precise analysis of agent-based model (ABM) outputs can be a challenging and even onerous endeavor. Multiple runs or Monte Carlo sampling of one's model (for the purposes of calibration, sensitivity or parameter-outcome analysis) often yield a large set of trajectories or state transitions which may, under certain measurements, characterize the model's behavior. These temporal state transitions can be represented as a directed graph (or network) which is then amenable to network analytic and graph theoretic measurements. Building on strategies of aggregating model outputs from multiple runs into graphs, we devise a temporally-constrained graph aggregating state changes from runs and examine its properties in order to characterize the behavior of a land-use change ABM, the RHEA model. Features of these graphs are transformed into measures of complexity which in turn vary with different parameter or experimental conditions. This approach provides insights into the model behavior beyond traditional statistical analysis. We find that increasing the complexity in our experimental conditions can ironically decrease the complexity in the model behavior.
|Title of host publication||Advances in Social Simulation 2015|
|Editors||W. Jager, R. Verbrugge, A. Flache, G. de Roo, L. Hoogduin, Ch. Hemelrijk|
|Place of Publication||Cham, Switzerland|
|Number of pages||10|
|Publication status||Published - 2016|