Nowadays, designing and building computer systems has become increasingly difficult; this is essentially due to the great number of existing solutions. This article proposes a hybrid heuristic allowing the connection between meta-models of different systems, which will allow the generation between models conforming to these connected meta-models. First, this article presents the architecture of the generative matching approach named generative automatic matching (GAM), then is introduced an important part of this approach, a hybrid heuristic allowing the matching between the meta-models. Finally, the authors conclude by a multiple criteria evaluation of this approach.