8.1.2. K-th Neighbour Cluster Analysis
K-th Neighbour clustering is also a hierarchical method, however, the basic algorithm is modified to overcome the inherent problem of noise associated with Single Linkage or Complete Linkage methods. See Wong, M A and Lane, T A (1983), pp. 362-8. Instead of linking a point to the nearest (or farthest) point, this method will link a point to its kth neighbour, where k is supplied by the user. However, since by symmetry this principle is also applied to the point to be linked, the distance from each point to its kth neighbour is computed. If the actual distance between the two points is greater than the sum total of their distances to their own kth neighbours, then the distance between the two points is assumed to be infinity (a very large number). Otherwise, the distance is the average of distances to the kth nearest points.
Output options for this procedure are the same as for other hierarchical methods.