Publication
Post--Mining on Association Rule Bases
Dietmar Seipel; Marcel Waleska; Daniel Weidner; Sven Rausch; Martin Atzmueller
In: Proc. European Conference on Artificial Intelligence Workshops. European Conference on Artificial Intelligence (ECAI-2023), Pages 23-35, Springer, 2024.
Abstract
Association rule mining identifies potentially unknown correlations between columns in a relational database and is therefore a central task in the data mining process. In many cases though, the resulting set of association rules is by far too big and confusing for a domain expert to extract useful knowledge. Moreover, the domain expert might often expect that particular association rules are generated by the data mining process and would like to be able to search, e. g., for similar rules according to their expectation.
In this paper, we propose to store association rules in a rule base which will offer functionality for the management of the rules. The rule base can also offer subsumption features for condensing huge sets of derived association rules to smaller subsets of interesting rules that can be investigated by the domain expert. The rule base can store association rules derived in different attempts of rule mining. And depending on the result of a rule mining attempt, further attempts can be initiated. The post–processing of the rule mining result, for example, can derive further association rules by grouping the values of attributes in their consequences; it can also initiate another round of pre-processing of the base relation to group certain values of certain attributes to sets (tiles), and thus to derive correlations between sets of attribute values in further steps of association rule mining.