Usage Des Mesures Pour La G\'en\'eration Des R\`egles d'Associations Cycliques

TL;DR

This paper introduces a novel method for extracting cyclic association rules from measures in data warehouses, enhancing the analysis of temporal patterns by redefining evaluation metrics and integrating aggregation functions.

Contribution

The paper presents a new approach to generate cyclic association rules from measures, considering multidimensional context and temporal summarizability, which was not addressed in prior work.

Findings

Effective extraction of cyclic association rules demonstrated on real data warehouse

Improved evaluation metrics for rule quality considering temporal aspects

Enhanced detection of regular temporal patterns in multidimensional data

Abstract

The online analytical processing (OLAP) does not provide any explanation of correlations discovered between data. Thus, the coupling of OLAP and data mining, especially association rules, is considered as an efficient solution to this problem. In this context, we mainly focus on a particular class of association rules which is the cyclic association rules. These rules aimed to discover patterns that display regular variation over user-defined intervals. Generally,the generated patterns do not take an advantage from the specificities of the multidimensional context namely, the consideration of the measures and their aggregations. In this paper, we introduce a novel method for extracting cyclic association rules from measures, and we redefine the evaluation metrics of association rules quality inspired of the temporal summarizability of measures concept through the integration of appropriate aggregation functions. To prove the usefulness of our approach, we conduct an empirical study on a real data warehouse.