Cure: an efficient clustering algorithm for large databases

A Maximum Profit Coverage Algorithm with Application to Small Molecules Cluster Identification

“…When clustering of data is considered, the intuitive definition of outliers becomes "points which do not belong to any of the clusters". A more specific definition is derived from the clustering target or the clustering process [4,6,7,10,16,27,28]. In the case of the CIM, the outliers are objects in a neighborhood not dense enough, according to the given definition of density.…”

“…For k-median outliers were considered by demanding payment on unclustered points and for k-center the number of points which are considered as outliers was added to the model. Heuristics for clustering data containing outliers were introduced, among others, in [4,6,7,10,16,27,28]. In hierarchical methods clusters that "grow" slowly are considered as outliers, whereas in the k-means method points that are far from all means are considered as outliers.…”

Landscape of clustering algorithms

“…• Four graph partitioning criteria similar to the CURE algorithm as described in [11], but with the above mentioned distance definitions (14-17).…”

“…Also, some clustering algorithms do not have explicit objective functions. Examples include mean-shift clustering [13] and CURE [11]. However, there is still the notion of optimality in these algorithms and they possess their objective functions, albeit defined implicitly.…”

A Comparative Study of Multi-SOM Algorithms for Determining the Optimal Number of Clusters

“…Several hierarchical clustering methods have been proposed such as: CURE [4], BIRCH [5], and CHAMELEON [6].…”