Efficiently Clustering Probabilistic Data Streams

Abstract

Data mining on uncertain data stream has attracted a lot of attentions because of the widely existed imprecise data generated from a variety of streaming applications in recent years. The main challenge of mining uncertain data streams stems from the strict space and time requirements of processing arriving tuples in high-speed. When new tuples arrive, the number of the possible world instances will increase exponentially related to the volume of the data stream. As one of the most important mining task, how to devise clustering algorithms has been studied intensively on deterministic data streams, whereas the work on the uncertain data streams still remains rare. This paper proposes a novel solution for clustering on uncertain data streams in point probability model, where the existence of each tuple is uncertain. Detailed analysis and the thorough experimental reports both on synthetic and real data sets illustrate the advantages of our new method in terms of effectiveness and efficiency.

This work is supported by Shanghai Leading Academic Discipline Project (Project Number: B412) and National Natural Science Foundation of China (NSFC) under grant No. 60803020.