Journal
JOURNAL OF SUPERCOMPUTING
Volume 76, Issue 12, Pages 10288-10318Publisher
SPRINGER
DOI: 10.1007/s11227-020-03247-5
Keywords
High average-utility itemset mining; Pruning methods; Upper bound; Utility list structure; Data mining
Categories
Funding
- Indian Institute of Technology (ISM), Dhanbad, Govt. of India
- Department of Computer Science & Engineering, Indian Institute of Technology (ISM), Dhanbad, India
Ask authors/readers for more resources
High-utility itemset mining is a prominent data-mining technique where the profit or weight of itemsets plays a crucial role in defining meaningful patterns. High average-utility itemset (HAUI) mining is an advancement over high-utility itemset mining, which introduces an unbiased measure called average utility to associate the utility of itemsets with their length. Several existing HAUI mining algorithms use various upper bounds such as average-utility upper bound, revised tighter upper bound, and looser upper bound to preserve pruning methods. However, these upper bounds overestimate the average-utility of itemsets and slow down the mining process. This paper presents a fast high average-utility itemset miner (FHAIM) algorithm, which uses two improved upper bounds and several efficient pruning strategies to avoid the processing of unpromising candidate itemsets. Moreover, a novel list structure named recommended average-utility list (RAUL) is presented to store the average-utility and the required information for pruning. The RAUL for an itemset can be constructed by joining the RAULs of its subsets to avoid excessive database scans. We have performed substantial experiments on various benchmark datasets to evaluate the performance of the FHAIM in comparison with two existing HAUI mining algorithms. Experimental results show that FHAIM outperforms the existing HAUI mining algorithms in terms of runtime, memory usage, join counts, and scalability.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available