Hits:

Indexed by:期刊论文

Date of Publication:2013-12-01

Journal:Journal of Computational Information Systems

Included Journals:EI、Scopus

Volume:9

Issue:23

Page Number:9261-9268

ISSN No.:15539105

Abstract:Probabilistic frequent itemset mining plays a significant role in association rule mining over uncertain data, which means to obtain itemsets with existential probability larger than a minimum threshold. So far, several studies have focused on this kind of work, however, two problems brought to users are: how to select an appropriate threshold for the mining and how to reflect their concern to different items. Against these issues, in this paper, we provide a framework for mining k weight-based probabilistic itemsets that user expects most. The concept of item weight is newly added to uncertain data mining, and k refers to the number of itemsets that user desires. Then algorithm tkWUG is proposed as an extension of the traditional UF-Growth. Moreover, a strategy is provided to solve the challenges brought by the new framework. Results on real and synthetic datasets show and demonstrate the accurancy and efficiency of new algorithm. Copyright ? 2013 Binary Information Press.