Abstract: Aiming at the problem that the partition of the item domain was not considered in the definition of regular association rules, we proposed that maximal association rules were useful complement to regular association rules. Using soft set theory and logical formulas on soft sets as the main tool, we solved the problem of the description of related core concepts for mining both regular and maximal association rules, and obtained unified mathematical characterization of these concepts. The example analysis shows that, compared with classical mining methods, maximal association rule mining method based on logical formulas on soft sets can effectively reduce redundancy and eliminate invalid rules, which improves the accuracy and mining efficiency of rule.

Key words: data mining, implementation set, association rule, logical formula, soft set