Utility Mining Algorithms - A Comparative Study

Utility mining is an emerging topic in data mining. The aim of utility mining is to discover the itemsets that have maximum utilities. Here utility refers number of items bought, cost of an item or it can be any other user choice in a transaction database. Frequent itemset mining is starting point of utility mining. In frequent itemset mining most often occurring itemsets in a transaction are retrieved. The discovery of such frequent itemsets can help in many business decision making process. Frequent itemset mining concentrates on the number of occurrence of items in a transaction, but not the value of items. But utility mining considers importance of itemsets like the profit it earns in a transaction, quantity in a transaction. In this paper various utility mining algorithms like MEU (Mining with expected utility), FUM (Fast Utility Mining), Two-Phase, CTU-Mine, UP-Growth (Utility Pattern Growth), and FHM (Faster High Utility itemset Mining) MHUI-BIT (Mining High-Utility Itemsets based on BIT vector), MHUT-TID (Mining High-Utility Itemsets based on TIDlist), and THUI (Temporal High Utility Itemsets) are discussed.