Fast pattern matching in conceptual models – evaluating and extending a generic approach

Abstract

Identifying structural patterns in conceptual models serves a variety of purposes ranging from model comparison to model integration and exploration. Although there are a multitude of different approaches for particular modelling languages and application scenarios, the modelling community lacks an integrated approach suitable for conceptual models of arbitrary languages and domains. Therefore, a generic set-theory based pattern matching approach has recently been developed. To prove that this approach is beneficial in terms of performance, we conduct a statistically rigorous analysis of its runtime behaviour. We augment the original approach to include a caching mechanism that further increases performance. We are able to show that the original algorithm is able to identify arbitrary patterns within milliseconds. The caching extension further increases performance by up to fifty per cent given the model base and patterns we used.