Mining flexible association rules from XML

The role of the eXtensible Markup Language (XML) is becoming very important in the research fields focusing on the representation, the exchange, and the integration of information coming from different data sources and containing information related to various contexts such as, for example, medical and biological data. Extracting knowledge from XML datasets is an important issue that may be difficult because of the semistructured intrinsic nature of XML; indeed documents can have an implicit and irregular structure, not defined in advance. In this paper, we propose a novel approach for discovering frequent, but approximate, information in XML documents, based on Flexible Tree Rules taking into account both structure and content of the analyzed data. Our proposal is flexible enough to be adapted to both documents with a regular structure and documents with a highly heterogeneous structure, and can be used to evaluate the similarity of XML documents. Moreover, we describe an algorithm to evaluate the similarity degree of a Flexible Tree Rule with respect to an XML document.