Type inference in flexible model-driven engineering

Abstract

In Model-Driven Engineering (MDE), models conform to metamodels. In flexible modelling, engineers construct example models with free-form drawing tools; these examples may later need to conform to a metamodel. Flexible modelling can lead to errors: drawn elements that should represent the same domain concept could instantiate different types; other drawn elements could be left untyped. We propose a novel type inference approach to calculating types from example models, based on the Classification and Regression Trees (CART) algorithm. We describe the approach and evaluate it on a number of randomly generated models, considering the accuracy and precision of the resultant classifications. Experimental results suggest that on average 80% of element types are correctly identified. In addition, the results reveal a correlation between the accuracy and the ratio of known-to-unknown types in a model.