Incorporating Deviation-Detection Functionality into the OLAP Paradigm

Although the integration of data mining and OLAP is obviously of great interest (if only due to the widespread availability of commercial-strength OLAP tools), relatively little progress has been made in this integration. The main goal of this paper is to contribute to this integration, by incorporating some data mining functionality (more precisely, deviation-detection functionality) into the OLAP paradigm. The data mining method addressed in this paper is Attribute Focusing. This method discovers interesting attribute values, in the sense that the observed frequency of those values deviate from their expected frequency. However, standard Attribute Focusing assumes that attribute values are “flat” (i.e. non-hierarchical), which is a serious limitation for its application in the OLAP paradigm. In order to make Attribute Focusing useful for OLAP we have adapted this method to the kind of hierarchical dimension often found in data cubes.