A novel approach for high utility closed itemset mining with transaction splitting

Abstract

Data mining techniques automate the process of finding decisional information in large databases. Recently, there has been a growing interest in designing high utility itemset mining algorithms. Utility pattern growth algorithm is one of the most fundamental utility itemset mining algorithms without candidate set generation. In this paper, a different possibility of designing a lossless high utility itemset mining algorithm is discussed. This algorithm can achieve high utility closed itemsets (HUCI’s) even when any number of long transactions present in the database. HUCI’s are generated by applying both UP Growth and Apriori concepts with closed itemset mining algorithm on the large database. Too many long transactions in database may affect the efficiency of the algorithm. To resolve this transaction splitting is used. This algorithm adopts transaction weighted downward closure property which guarantees only promising items are high utility items. The proposed algorithm will generate high utility closed itemsets in an efficient way.