CURE (Clustering Using REpresentatives) is an efficient data clustering algorithm for large databases. Compared with K-means clustering it is more robust to outliers and able to identify clusters having non-spherical shapes and size variances.