Abstract:
High utility itemsets mining (HUIM) is an interesting topic in data mining which can be applied in a wide range of applications, for example on retail marketing to find sets of sold products that give high profit, low cost, etc. However, HUIM only considers utility values of items/ itemsets which may be insufficient to observe buying behavior of customers. To address this issue, we here introduce an approach on pushing regularity constraint on high utility itemsets mining to observe occurrence behavior of high utility itemsets. Based on this approach, sets of co-occurrence items with (i) high utility values and (ii) regular occurrence, called high utility-regular itemsets (HURIs), are regarded as interesting. To mine HURIs, an efficient singlepass algorithm, called HURI-UL, is proposed. HURI-UL applies the concept of remaining and overestimated utilities of itemsets to early prune search space and also utilizes utility list structure to efficiently maintain utility values and occurrence information of itemsets. Experimental results on real datasets show that our proposed approach is efficient to discover high utility itemsets with regular occurrence.