A FAST ALGORITHM FOR MINING K-ITEM HIGH UTILITY ITEMSETS

Abstract

High utility itemset mining performs to find sold item sets whose profit overcomes a given threshold. Finding high utility itemsets helps to suggest related items on e-commerce sites and makes effective policies of sales. Recommender systems usually introduct from 5 to 7 related products to help users choose products. In previous studies, finding high utility itemsets often consume the time because many combinations of items were considered. In this paper, we present a fast algorithm for mining k-item high utility itemsets. A compact list structure is used to store information about the itemsets appearing from the curent database. First, the proposed algorithm perform a vertical segmentation of the transaction database to obtain sub-partitions. Next, mine high utility itemsets on each subpartiton by listing k-itemsets and its utility. Experiments were performed on brenchmark databases. Experimental results show that the proposed approach significantly reduces both memory and computation time. Moreover, it is better than the compared algorithm.