Distance-Based Outlier Detection on Uncertain Data

The technique of outlier detection is useful in many real world applications such as detection of network intrusion. It has been studied intensively on deterministic data. However, it is still a novel research field on uncertain data. To our best knowledge, this paper is the first one to focus on distance-based outlier detection on uncertain data, in which each data is affiliated with a certain confidence value. In this paper, we propose a new definition of outlier on uncertain data. Based on the properties we discovered, both dynamic programming approach (DPA) and grid-based pruning approach (GPA) are used for detecting outliers on uncertain data efficiently. Detailed analysis and thorough experimental results demonstrate the efficiency and scalability of our method.