Abstract for:Efficient Parameter Exploration and Behavior Mapping for System Dynamics Models

System dynamics (SD) models are widely used for studying dynamic phenomena that can be characterized by their nonlinear and feedback-rich nature in their behavior space. Exploring this space to identify plausible modes of model behavior is of interest to analysts at several stages of the modeling cycle. One of the major obstacles for extensive behavior space is about efficient exploration of the parameter space. Without any apriori knowledge, an analyst needs to explore the space either by a systematic brute force search, or randomly. In both cases, developing a good understanding about the behavior modes that can be observed in a given parameter space requires an extensive set of parameter evaluations. In this study we propose an approach that efficiently explores and pictures the behavior space of a given model with minimal loss in precision for any number of dimensions. Suggested model is demonstrated by constructing a behavior map of an SD model based only on 300 parameters and compared with the one developed by brute force enumeration of 10000 parameters. It is shown that proposed approach constructs the behavior map with 90\% accuracy in 20\% of the time and corrects some of the errors made by the classification algorithms.