An XML Format for Association Rule Models based on the GUHA Method Tom´ aˇ s Kliegr, Jan Rauch University of Economics, Prague, Dept. Information and Knowledge Engineering, N´ am. Winstona Churchilla 4, 130 67 Praha 3, Czech Republic, tomas.kliegr@vse.cz|rauch@vse.cz Abstract. This paper proposes the GUHA AR Model, an XML Schema- based formalism for representing the setting and results of association rule (AR) mining tasks. In contrast to the item-based representation of the PMML 4.0 AssociationModel, the proposed expresses the association rule as a couple of general boolean attributes related by condition on one or more arbitrary interest measures. This makes the GUHA AR Model suitable also for other than apriori-based AR mining algorithms, such as those mining for disjunctive or negative ARs. In addition, there are prac- tically important research results on special logical calculi formulas which correspond to such association rules. The GUHA AR Model is intended as a replacement of the PMML AssociationModel. It is tightly linked to the Background Knowledge Exchange Format (BKEF), an XML schema proposed for representation of data-mining related domain knowledge and to the AR Data Mining Ontology ARON. 1 Introduction In recent years, the advent of service oriented data mining have been decreasing the costs of advanced analysis for the end user, while semantic technologies and stronger orientation towards the web have opened up new possibilities for post-processing and sharing data mining results. At the same time, the pace of innovations in data mining algorithms is increasing. In this situation, a strong need arose for a generally accepted and maintained standard for exchange of mining models, which was filled by the XML-based Predictive Model Markup Language (PMML) from the DMG consortium. Its latest version 4.0 released in June 2009 supports twelve types of mining models, including association rules. Although PMML is a widely accepted standard, its AssociationModel (fur- ther AR Model) as of its current version 4.0 lacks support for new types of association rule mining algorithms stifling their deployment into industry. For example, the PMML AR Model does not standardize representation of disjunc- tive association rules (e.g. [14, 15]), global constraints as needed e.g. in local mining of association rules [16], new interest measures [23] and constraints in- volving multiple arbitrary interest measures as e.g. used in [5]. In this paper, we suggest for discussion in the data mining community a new format for representation of association models that covers features mentioned