keyboard_arrow_up
Nuclear: An efficient method for mining frequent itemsets based on kernels and extendable sets

Authors

Huy Quang Pham1,2, Duc Tran3, Ninh Bao Duong2, Philippe Fournier-Viger4, Alioune Ngom1, 1University of Windsor, Windsor, Canada, 2University of Dalat, Vietnam, 3Ho Chi Minh City University of Food Industry, Vietnam and 4Harbin Institute of Technology Shenzhen Graduate School, China

Abstract

Frequent itemset (FI) mining is an interesting data mining task. Directly mining the FIs from data often requires lots of time and memory, and should be avoided in many cases. A more preferred approach is to mine only the frequent closed itemsets (FCIs) first and then extract the FIs for each FCI because the number of FCIs is usually much less than that of the FIs. However, some algorithms require the generators for each FCI to extract the FIs, leading to an extra cost. In this paper, based on the concepts of “kernel set” and “extendable set”, we introduce the NUCLEAR algorithm which easily and quickly induces the FIs from the lattice of FCIs without the need of the generators. Experimental results showed that NUCLEAR is effective as compared to previous studies, especially, the time for extracting the FIs is usually much smaller than that for mining the FCIs.

Keywords

Association Rule, Kernel and Extendable Set, Frequent Itemset, Frequent Closed Itemset, Nuclear

Full Text  Volume 9, Number 6