An Approach to Cluster Uncertain Data Objects Using Voronoi Diagram and Indexing

Recently, the clustering of uncertain data became a hot topic in data mining since more and more applications, such as sensor database, location database, biometric information systems produce vague and imprecise data. Though, there exist lots of approaches to classify the uncertain data by hard classifiers, few of them address the classification of the uncertain data by soft classifiers. In this approach,apply K-Means algorithm to generate the clusters. It is an unsupervised clustering algorithm. It uses simple and easy way to cluster the dataset. This clustering algorithm uses the Expected Distance (ED) to compute the distance between objects and cluster representatives. The expected distance calculations are the performance bottleneck of the algorithms. Also it is an iterative algorithm which generate clusters until the centroids no longer moves. To improve the performance of K-Means, there has been tried an effective technique called Voronoi Diagrams from Computational Geometry. This technique works efficiently but provides erroneous uncertain clustering and takes longer time. To increase time efficiency effective indexing must be integrated in this. To incorporate indexing we introduced a naive indexing technique over these uncertain data objects, so that it reduces the computational cost. Hence, approach of integrating K-Means, Voronoi diagram and indexing applied over uncertain data objects generates imposing outcome when compared with the accessible methods. Therefore proposed approaches can efficiently optimize clustering algorithm, reduce the corresponding searching space and improve the performance on uncertain data clustering.