| Popis: |
As a hot topic of association rule mining, high average utility itemset mining (HAUIM) provides a scientific measurement of an itemset’s utility, thus it can be adopted to find out the rules that are critically useful to the real applications. However, there is a prominent deficiencty in existing algorithms for HAUIM, a huge number of memory consumption are needed during the generation of high average utility itemset. Hence, an efficient algorithm called BAUL-Miner is proposed to solve the memory problem, in which a buffered average utility list with a more compact upper bound and two pruning strategies are introduced to reduce the searching space. Finally, detail experiments in synthetic and industrial datasets verify the effectiveness of BAUL-Miner, it outperforms traditional HAUIM algorithms in runtimes, memory consumption and scalability. |
