An Efficient Search Space Exploration Technique for High Utility Itemset Mining

Abstract

Listen

Translate article

Data mining techniques discover the potentially helpful pattern concealed in a vast database for the decision support system in various real-life applications. One such technique is association rule mining. It finds the pattern from the transaction database to understand customer behavior. Frequent itemset mining (FIM) identifies a set of frequently purchased items together. One key drawback of FIM is that it ignores the item's importance. An item's importance plays a pivotal role in a real-world application. Therefore, it is necessary to discover the essential itemset from the transaction database that generate the high profit called the HUIM (High-Utility Itemset Mining) problem. From a transaction database, a variety of strategies can be used to find high utility itemset. The HUIM techniques that use the Utility list are recent and have better performance in terms of memory utilization and running time. The key limitation of these algorithms is performing costly utility list join operations. This paper proposes an efficient search space exploration technique to lessen the number of comparisons performed for utility list join operations. Hence, it reduces the cost of utility list join operations. In addition, the execution time of the proposed method is evaluated. Further, the same is compared with appropriate existing state-of-the-art methods. Finally, extensive experiments carried out on publicly available benchmark datasets show that the proposed support count ascending order (SCAO) based approach performs much better as oblivious to the state-of-the-art methods.