Clustering was done based on the affinity between the data points to cluster the data. Objects within the cluster will have the Euclidean distance very low or closely related to each other whereas distance between the data points in separate clusters are high or not closely related. we uses the word Exemplar, it is similar to the data points but additionally it acts like a role model to the data points usually we select center as an exemplar. After choosing exemplar, we need to be very careful as we need to iteratively refine it until it works fine better. Choosing bad exemplar will fiasco the output. Real valued messages are exchanged between the data points till a perfect exemplars and clusters are created.