Title: Mining convergent and divergent sequences in multidimensional data

Authors: M. Plantevit, A. Laurent, M. Teisseire

Addresses: GREYC, CNRS – UMR 6072, Universite de Caen, Campus Cote de Nacre, F-14032 Caen Cedex, France. ' LIRMM, CNRS – UMR 5506, Universite Montpellier 2, 161 Rue Ada, 34392 Montpellier Cedex 5, France. ' LIRMM, CNRS – UMR 5506, Universite Montpellier 2, 161 Rue Ada, 34392 Montpellier Cedex 5, France

Abstract: Online Analytical Processing (OLAP) mining provides useful knowledge to decision makers from multidimensional data stored in data warehouses. However, it is still difficult to find data mining tools taking all the data specificities (e.g., multidimensionality, hierarchies, time) into account. In this paper, we propose an original method to discover multidimensional sequential patterns among several levels of hierarchies. We define two types of multidimensional sequences: convergent sequences, where the elements become more and more precise, and divergent sequences where the elements become more and more general. A pattern-growth based algorithm is proposed and is shown to be efficient in our experiments.

Keywords: data mining; sequential patterns; multidimensional data; hierarchies; online analytical processing; OLAP mining.

DOI: 10.1504/IJBIDM.2009.029074

International Journal of Business Intelligence and Data Mining, 2009 Vol.4 No.3/4, pp.242 - 266

Published online: 03 Nov 2009 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article