The TwoKey Plot for Multiple Association Rules Control

Abstract

Association rules have been popular in theory, though it is unclear how much success they have had in practice. Very many association rules are found in any application by any approach and they require effective pruning and filtering. There has been much research in this area recently, but less with the goal of providing a global overview and summary of all rules, which may then be used to explore the rules and to evaluate their worth. The unusual feature of association rules is that those with the highest objective values for the two key criteria (support and confidence) are not usually those with the most subjective interest (because we know the obvious results already). The TwoKey plot is a way of displaying all discovered association rules at once, while also providing the means to review and manage them. It is a powerful tool in order to get a first overview of the distribution of confidence and support. Features such as separate groups of rules or outliers are detected immediately. By exploiting various ancestor relationship structures among the rules, we can use the TwoKey Plot also as a visual assessment tool, closely related to pruning methods — e.g. those proposed by Bing Liu (1999). The concept will be illustrated using the interactive software MARC(Multiple Association Rules Control).

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.