Relationships between objects constitute an important component of domain knowledge and represent an added value to data and knowledge discovery and management. Some existing studies in data mining (including association rule mining) exploit domain knowledge to better conduct tasks in applications such as e-learning and e-commerce. However, general semantic links between objects are either not exploited in the mining process or limited to is-a relations.

In this paper we present the basis of a new approach where direct and indirect domain relations between items (basket market analysis) are used to reveal hidden knowledge in item transactions. First, we define the notion of relation association rule as an association that holds between sequences of relations. Such a type of rules is produced from item transactions and existing links among items in a given domain ontology. Then, we define an a priori-based algorithm for relation association rule mining. Finally, we illustrate the potential of our approach for recommendation purposes.