A Framework for High-Quality Clustering Uncertain Data Stream over Sliding Windows

Abstract

In recent years, data mining over uncertain data stream has attracted a lot of attentions along with the imprecise data widely generated. In many cases, the estimated error of the data stream is available. The estimated error is very useful for the clustering process, since it can be used to improve the quality of the cluster results. In this paper, we try to resolve the problem of clustering uncertain data stream over sliding windows. The tuple expected value and uncertainty are considered meanwhile in the clustering process. We therefore propose the algorithm based on Voronoi diagram to reduce the number of expected distance calculation over sliding windows. Finally, our performance study with both real and synthetic data sets demonstrates the efficiency and effectiveness of our proposed method.

This research was supported by the National Natural Science Foundation of China (Grant No. 61073063, 61173029, 60803026 and 61173030).