An Ontology-based Framework for Mining Patterns in the Presence of Background Knowledge

Since its formulation, pattern mining has been focused both in the development of mining algorithms and in its extension to address new challenges, like the discovery of structured patterns. In both cases, existing approaches are now effective and efficient. However, and despite pattern mining is a constrained problem by definition, there is no generally accepted solution to incorporate domain knowledge into the mining process. In this paper, we propose the Onto4AR framework and explain why it can be that solution. The framework uses a domain ontology to represent background knowledge, that can be used to impose constraints in the mining process. The user benefits from a set of pre-defined constraints, specified over general relations among concepts, like is-a and has-a relations. These constraints provide a mining environment independent of the problem domain, which can be extended with the definition of new constraints, either general (temporal or structural, for instances) or domain dependent. The framework’s potential is illustrated with its application in a case study.