Enumerating Minimal Generators from Closed Itemsets – Toward Effective Compression of Negative Association Rules
Koji Iwanuma, Kento Yajima, Yoshitaka Yamamoto · 2021 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE) · 2021
Negative association rules are valuable and essential for expressing various latent properties which hide in big data. The number of valid negative association rules, however, always becomes so huge, thus an effective compression method of the set of negative rules is quite important. Minimal generators are very useful for compressing the set of valid negative rules. In this paper, we study several efficient algorithms for enumerating minimal generators from given closed itemsets. Especially we propose a novel enumeration algorithm which does not use any support computation, but uses an eager hash search in a top-down tree search. We show experimental results for evaluating our enumeration algorithms, and confirm very good performance of the enumeration method without support computation.