Des blocs de données aux motifs graduels multidimensionnels

Coupling data mining and data warehousing allows for discovering relevant information from data cubes. In this framework, several methods have been proposed, aiming for instance at discovering association rules or sequential patterns. However, no method has been proposed to discover gradual rules from such multidimensional databases. In this paper, we thus propose to discover correlations between a set of ordered dimensions with the measure evolution. Such rules can be achieved using a summary of data cube called blocks. We first describe a new algorithm for the extraction of such blocks, and then an efficient algorithm to extract gradualness from these blocks.