Distance-Based Outlier Detection on Uncertain Data of Gaussian Distribution

Abstract

Managing and mining uncertain data is becoming important with the increase in the use of devices responsible for generating uncertain data, for example sensors, RFIDs, etc. In this paper, we extend the notion of distance-based outliers for uncertain data. To the best of our knowledge, this is the first work on distance-based outlier detection on uncertain data of Gaussian distribution. Since the distance function for Gaussian distributed objects is very costly to compute, we propose a cell-based approach to accelerate the computation. Experimental evaluations of both synthetic and real data demonstrate effectiveness of our proposed approach.