Mining frequent itemsets from sparse data streams in limited memory environments

Abstract

Floods of data can be produced in many applications such as Web click streams or wireless sensor networks. Hence, algorithms for mining frequent itemsets from data streams are in demand. Many existing stream mining algorithms capture important streaming data and assume that the captured data can fit into main memory. However, problem arose when the available memory so limited that such an assumption does not hold. In this paper, we present a data structure called DSTable to capture important data from the streams onto the disk. The DSTable can be easily maintained and is applicable for mining frequent itemsets from streams (especially sparse data) in limited memory environments. © 2013 Springer-Verlag Berlin Heidelberg.