A Sliding-Window Approach for Finding Top-k Frequent Itemsets from Uncertain Streams

Abstract

The analysis and management of uncertain data has attracted a lot of attention recently in many important applications such as pattern recognition and sensor network. Frequent itemset mining is often useful in analyzing uncertain data in those applications. However, previous works just focus on the static uncertain data instead of uncertain streams. In this paper, we study the problem of mining top-k FIs in uncertain streams. We propose an efficient algorithm, called UTK-FI, based on sliding-window and Chernoff bound techniques for finding k most frequent itemsets of different sizes. Experimental results show that our algorithm performs much better than many established methods in uncertain streams environment.