J. Han, M. Kamber, and A. K. H. Tung. Geographic Data Mining and Knowledge Discovery, chapter Spatial Clustering Methods in Data Mining: A Survey, pages 1–29. Taylor and Francis, 2001. [url]
This paper presents a review of spatial clustering methods, which are considered an important component of spatial data mining. The authors classify the methods into four categories: partitioning method, hierarchical method, density based method, and grid-based method.
Partitioning methods like the k-means , the k-medoids and EM clustering are methods which make uses of a techinque called iterative reallocation to improve the clustering quality from an initial solution. These methods tend to find clusters that are of sperical shape and they are made for minimising the disctance from the data objects to their distance centers.
On the contrary of these, hierarchical clustering algorithms fixed the membership of a data object once it has been allocated to a cluster. BIRCH, CURE and CHAMELEON uses complex criteria for compressing and relocating data before merging clusters.
A third group of these methods is based on density of data points within a region to discover clusters. Belong to this category methods like DBSCAN, OPTICS and DENCLUE.
Finally, to increase the efficiency of clustering, grid-based clustering methods approximate the dense regions of the clustering space by quantizing it into a finite number of cells that contain more than a number of points as dense. Clusters are then formed by connecting the dense cells. To this category belongs STING, and CLIQUE.