Seminář: Searching for the (golden) needle in the association rule haystack

The SEWEBAR project aims at post-processing association rule mining results in an attempt to address the key problem of Association Rule mining task: AR mining algorithms either output too many rules, most of them uninteresting to the domain expert (1), or few rules but with a high risk of omitting the really interesting ones (2). At the seminar, we will present an approach to tackling this problem by empowering the domain expert with an association rule search engine. Three different approaches to construct such a search engine will be shown and demonstrated: the information retrieval approach based on the Lucene project, the structured search (XML/XQUERY) approach based on the XML database Berkeley, and the semantic search (Topic Maps/tolog) approach utilizing Ontopia Knowledge Suite. Following the seminar, there will be an internal SEWEBAR meeting on the work-in-progress on the Everminer project. (Core slides are in English, students‘ slides are in Czech.)