A prototype cloud-based reproducible data analysis and visualization platform for outputs of agent-based models

Abstract

Agent-based models typically have stochastic elements and many potential parameter combinations. This requires that we conduct multiple model runs to sweep the parameter space, creating large quantities of computationally generated, hyper-dimensional, “big data”. Understanding the models’ implications requires structured exploration of these complex output data. In response to this need, the MIRACLE team has developed a prototype web application that enables researchers who archive their model output data and analysis methods to perform online output data exploration and reproducible, re-parameterizable data analysis. We plan to build on this prototype, integrating with broader reproducibility initiatives in scientific computation and big data, to facilitate improved communication within research groups, and increase access and transparency for external research community and the general public. This paper provides contextual background and a case study of the prototype MIRACLE data storage and analysis web tool.