Collective Intelligence for Model Transformation by Example

A transformation engine carries out the execution of the model transformation specification to generate the target model from its source model. This transformation represents a complex aspect of Model-Driven Engineering (MDE) and involves concepts specific to the technology. Nevertheless, the literature and practice on this topic define CASE tools that are used by designers as semi-automatic or automatic solutions. The use of CASE tools in software engineering poses a risk due to their automatic or semi-automatic recommendations for model creation, without direct designer oversight. This can lead to suboptimal or irrelevant models that do not align with project requirements. Designers must be aware of these limitations and supplement tool usage with their expertise to ensure the quality of generated models. In this context, designers are responsible for many tasks in the model transformation process, including selecting/dropping objects (e.g., attributes, relations), setting the number of configuration parameters, splitting, merging a structure, and disseminating the results. In order to fully model the transformation process, we must replace designers with a method that emulates the same expertise of designers and human intelligence during shared design, while leveraging collective intelligence. To do so, we propose using an Evolutionary Multi-Agent System (EMAS) called MeTeo-EMAS to emulate human actions in the transformation model that best self-optimizes/self-configures during model transformation, harnessing collective intelligence. We describe the components of our approach and show how MeTeo-EMAS agents use their goal model to reason about goals during their model transformation process. Finally, usage examples of MeTeo-EMAS are provided to highlight its importance in harnessing collective intelligence for effective model transformation.