Uncertain data cluster based on DBSCAN

Uncertain data mining has recently attracted interests from researchers due to its presence in many applications such as Global Positioning System (GPS) Wireless Sensor Networks (WSN), Moving Object Tracking. This paper is researching uncertain data clustering problem, almost all the existed algorithms of uncertain data calculate expectation to express the distance of objects, so they can cluster like certain data. But they neglect the distribution of objects and consume much more running time to calculate expectation. In the paper, we propose CIR-DBSCAN, an algorithm based on a representation model of distance distribution between uncertain objects, which uses the Core Influence Rate (CIR) to extend the traditional DBSCAN algorithm in uncertain data. To evaluate its performance and accuracy, a comparison against the clustering algorithm FDBSCAN is performed using synthetic datasets. The experimental results show that the proposed algorithm CIR-DBSCAN outperforms FDBSCAN in some cases.