Mining maximal frequent itemsets from Tuple-evolving data streams

Abstract

Today, most of the data mining applications exhibiting high data flow rate, and expecting algorithms to match the flow rate with redundant less knowledge. Data streaming applications consider every incoming transaction as a new tuple, irrespective of whether it is old tuple that gets revised or not. This kind of revision in data streaming application gives new and hidden knowledge, also brings new challenges and issues to the tasks. One of the issue is, interested/frequent itemsets may turn to infrequent or infrequent itemsets may turned into frequent, and other one is redundancy in output. In this paper, we address solution to the redundancy in output by finding maximal itemsets from tuple revision data streams. We propose SlideTree data structure to maintain stream data, and Lattice-Tree to maintain maximal itemset information. We propose an Update algorithm that combines effective data structures that derives all the Maximal itemsets over the tuple evolving data streams.