Finding class association rules (CARs) is one of the most important research topics in data mining and knowledge discovery, with numerous applications in many fields. However, existing techniques usually generate an extremely large number of results, which makes analysis difficult. In many applications, experts are interested in only the most relevant results. Therefore, we propose a method for querying top-k CARs based on their supports. From the set of mined CARs that satisfy the minimum support and the minimum confidence thresholds, we use a QuickSort-based method to query top-k rules. The whole rule set is partitioned into two groups. If the number of rules in the first group is k, then the first group is the set of result rules. If the number of rules in the first group is greater than k, the second group is partitioned to find the remaining top-k rules. Experimental results show that the proposed method is more efficient than existing techniques in terms of mining time.

data mining, class association rules, top-k class association rules