Effective Mining of High Utility Itemsets with Automated Minimum Utility Thresholds

Abstract

Utility mining is the recent data science task in the ground of data mining. Utility mining observes profit and quantity of each distinct item present in the transactions, thus it results in productive patterns with high importance in transactional databases. There are many algorithms sketched to trace the entire set of highly productive utility itemsets using user-defined single minimum utility threshold. An efficient framework called high utility itemset mining with automated minimum utility thresholds (HUIM-AMU) is put forward in this research paper. This algorithm uses a condensed tree arrangement called utility pattern tree to store the transactions and a constant value indicating the amount of most profitable itemsets. With very large count of transactions in a database, it is very difficult to identify the importance or productivity of every item. Without the knowledge of items, threshold setting may degrade the effectiveness of the process. In our proposed work, the difficulty in the setting of minimum threshold and the time spent on the analysis of database to set threshold are reduced by automating the threshold setting process. The results clearly indicate that the HUIM-AMU generates only profitable and compact itemsets.