Skip to Main Content
Frequent pattern mining is fundamental to many important data mining tasks. Many researchers had presented many mining methods in static database. Due to many special characters of data stream, those methods fail to be used in dynamic environment. We develop a novel method mining frequent items from data stream based on sliding window model. We use some compact data structures which make uses of the limited space efficiently. The proposed method is an approximate algorithm, it can eliminate the influence of old data to mined result. And the mined results are kept in a heap. This data structure is seldom used in other methods, and the mined results can be inquired by top-k.
Date of Conference: 26-28 Nov. 2007