Sensitivity analysis in model-driven engineering

Sensitivity analysis has been used in scientific research to explore the validity of models. Software engineering is inherently uncertain; we propose that sensitivity analysis can be used to analyse and quantify the effects of uncertainty when model management operations are applied to models. In this paper, we consider forms and measures of uncertainty in software engineering models. Focusing on data uncertainty, we present a framework for sensitivity analysis, and create an instantiation of the framework for the CATMOS decision-support tool. We show how this can be used to qualify the output of the entailed model management operations and thus improve both the confidence and understanding of models.