Density-Based Probabilistic Clustering of Uncertain Data

Abstract

In many applications like moving-objects and sensors databases, data values are inherently uncertain. In these systems, an attribute value can be modeled as a range of possible values, associated with a probability density function. Data mining of the uncertain data attracts more and more research interest recently. The definitions of probabilistic core object and probabilistic density-reachability are presented and a density-based probabilistic clustering algorithm for uncertain data is proposed, based on DBSCAN algorithm and probabilistic index on uncertain data. Simulation results show that the proposed algorithm outperforms other density-based clustering algorithm for uncertain data in accuracy and efficiency of clustering.