Towards Comprehensive Concept Description Based on Association Rules

The paper presents two approaches to post-processing of association rules that are used for concept description. The first approach is based on the idea of meta-learning; a subsequent association rule mining step is applied to the results of "standard" association rule mining. We thus obtain "rules about rules" that in a condensed form represent the knowledge found using association rules generated in the first step. The second approach finds a "core" part of the association rules that can be used to derive the confidence of every rule created in the first step. Again, the core part is substantially smaller than the set of all association rules. We experimentally evaluate the proposed methods on some benchmark data taken from the UCI repository. The system LISp-Miner has been used to carry out the experiments.