An Efficient Methodology for Clustering Uncertain Data Based on Similarity Measure

Abstract

In data mining clustering is one of the most important and crucial task. There exists various clustering techniques like partitioning and density based techniques. The previous methods worked on traditional partitioning methods which are found unreliable to handle uncertain data as uncertain objects are geometrically indistinguishable. Data uncertainty can be occurred in various data collecting systems like sensors, weather information collecting systems, etc.... Clustering gained much interest in clustering of uncertain data because the clustering algorithms works on the certain data so there is a need to do work on the algorithms to be as useful for uncertain data. The proposed work introduces the well-known Skew Divergence to measure the similarity between uncertain data. Previous methodology worked on KL-Divergence as well as on Janson-Shannon Divergence as the similarity measure. As KL-Divergence is very costly as compared to Skew Divergence or even infeasible. Skew Divergence with FCM creates a good combination and proves to be the best method of clustering uncertain data.