From the beginning of the data analysis system cluster computing plays an important role on it. The very early developed clustering algorithms which can handle only numerical data and K-means clustering is one of them and was proposed by Macqueen  in 1967. This algorithm helps us to find the homogeneity of the data set. This K-means algorithm has been modified in many ways to get the modified K-means and kernel based K-means is one of them. It is a nonlinear transformation which transforms the sample data into high dimensional feature space. Though this kernel based K-means performs good almost on every data set but it is unable to handle uncertainty. After rough set theory has been proposed by Pawlak , we have many clustering algorithms based on it which can handle uncertainty and heterogeneous data and Rough based K-means is one of them. So in this paper we are proposing the combination of these two methods and known as kernel based K-Means using rough set. © 2012 IEEE.