The use of different domain-specific modeling languages and diverse versions of the same modeling language often entails the need to translate models between the different languages and language versions. The first step in establishing a transformation between two languages is to find their corresponding concepts, i.e., finding correspondences between their metamodel elements. Although, metamodels use heterogeneous terminologies and structures, they often still describe similar language concepts. In this paper, we propose to combine structural metrics (e.g., number of properties per concept) and syntactic metrics to generate correspondences between metamodels. Because metamodel matching requires to cope with a huge search space of possible element combinations, we adapted a local and a global metaheuristic search algorithm to find the best set of correspondences between metamodels. The efficiency and effectiveness of our proposal is evaluated on different matching scenarios based on existing benchmarks. In addition, we compared our technique to state-of-the-art ontology matching and model matching approaches.