CLUSTERING ON UNCERTAIN DATA BASED PROBABILITY DISTRIBUTION SIMILARITY

Abstract

Clustering is important task in data mining. The main purpose of clustering is grouping the same object data
in a huge dataset and finding similarities between the objects. Clustering on unsure data is a most difficult task in both
modeling similarity between unsure data objects and producing efficient computational method. Clustering uncertain
data problems have been solved by using many different new data mining techniques and various algorithms. Techniques
have recently been suitable for clustering uncertain data based upon the traditional dividing clustering methods like kmeans and density-based clustering methods like DBSCAN to unsure data, they will determined by geometric distances
between objects. Computing the similarity between the data objects will be based upon a similarity distance measure and
further clustered with occurrence based clustering or hierarchical clustering methods. Such methods cannot handle
uncertain items that are geometrically no difference. In the proposed system we could using probability that are essential
characteristics of uncertain objects, and are considered in measuring likeness between uncertain objects. The very
popular technique Kullback-Leibler divergence used to procedures the distribution similarity between two uncertain data
items. First the probability division method for model unsure data object then there after measure the similarity between
data objects using distance metrics, then finally best clustering methods such as partition clustering, density clustering.