Pattern-based genetic model refinements in MDA

Abstract

We explore the application of genetic algorithms in model transformations that can be understood as pattern-based refinements. In MDA (Model Driven Architecture), such transformations can be exploited for deriving a PIM model from a elM model. The approach uses design patterns as the basis of mutations and exploits various quality metrics for deriving a fitness function. A genetic representation of models and their transformations is proposed, and the genetic transformation process is studied experimentally. The results suggest that genetic algorithms provide a feasible vehicle for model transformations, leading to convergent and reasonably fast transformation process. However, more work is needed to improve the quality of the individual models produced by the technique.