Efficient high utility itemset mining using buffered utility-lists

摘要: Discovering high utility itemsets in transaction databases is a key task for studying the behavior of customers. It consists finding groups items bought together that yield profit. Several algorithms have been proposed to mine using various approaches and more or less complex data structures. Among existing algorithms, one-phase employing utility-list structure shown be most efficient. In recent years, simplicity has led development numerous based tasks related mining. However, major limitation creating maintaining utility-lists are time consuming can consume huge amount memory. The reasons lists built intersection/join operation construct costly. This paper addresses this issue by proposing an improved called buffer reduce memory consumption speed up join operation. integrated into novel algorithm named ULB-Miner (Utility-List Buffer itemset Miner), which introduces several new ideas efficiently discover itemsets. uses designed store retrieve utility-lists, reuse during mining process. Moreover, also linear method constructing segments buffer. An extensive experimental study on datasets shows relying highly efficient terms both execution consumption. 10 times faster than FHM HUI-Miner consumes 6 it performs well dense sparse datasets.