Improving the discovery of association rules with intensity of implication

Abstract

In this paper, we propose a way to improve the rule-learning step in a Knowledge Discovery in Databases (KDD) process. Our purpose is to make possible the discovery of relevant rules in a large database. To achieve this goal, we merge : • a quality index proposed by R. Gras : intensity of implication, • together with a specific algorithm written by Agrawal et al. The algorithm itself is efficient in a large database but delivers a prohibitively large set of knowledge. Intensity of implication is a new measurement of the quality of association rules. Hence, we analyze it in detail and compare it with conditional probability index. We show that it is possible to significantly improve the relevance of association rules supplied by the algorithm proposed by Agrawal et al, by using the quality index : intensity of implication. An improved algorithm has been implemented, and has been tested both at the experimental level and on a real-life database.