A Framework for Fast Graph-Based Pattern Matching in Conceptual Models

Abstract

We introduce a pattern matching approach for conceptual models suitable for a number of model analysis scenarios like process weakness detection, process compliance checking, syntax verification and model translation. The approach does not depend on any particular modeling language which is achieved by treating conceptual models as labeled graphs. Consequently, we use pattern matching techniques known from algorithmic graph theory - sub graph isomorphism and sub graph homeomorphism. In general, algorithms solving these problems can be computationally expensive. However, special properties of conceptual models such as low tree width and planarity can be exploited to keep computational complexity manageable. This makes pattern matching applicable even to large models typically used in large companies or corporate groups. We introduce a high-level meta algorithm checking structural properties of input models and patterns to decide which low-level pattern matching algorithm will likely deliver search results quickest.