Profit and Quantity Oriented Two Efficient Approaches for Utility Pattern Mining

Traditional methods of association rule mining consider the appearance of an item in a transaction, whether or not it is purchased, as a binary variable. But, the quantity of an item purchased by the customers may be more than one, and the unit cost may not be the same for all items. A generalized form of the share mining model introduced to overcome this problem is utility mining. Developing an efficient algorithm is vital for utility mining because high utility itemsets cannot be identified by the pruning strategy. In this paper, we present two efficient approaches for utility pattern mining with the aid of FP-growth algorithm. The efficiency of utility pattern mining is achieved with two major concepts: 1) Incorporating the utility values after mining the frequent patterns (IUA-FP). Here, the patterns that are mined from the FP-growth algorithm are utilized to generate high utility patterns using internal and external utility. 2) Incorporating the utility values before mining the frequent patterns (IUB-FP). At this point, individual items that are less significant are taken out from the input database by considering their frequency along with their internal and external utility. Then, we apply the FP-growth algorithm in the transformed database to mine high utility patterns. Experimentation is carried out on these two concepts using synthetic dataset, T10I4D100K, attained from the IBM dataset generator and the performance study shows that the proposed two approaches are efficient in mining high utility patterns.