A Hybrid Clustering Algorithm for Identifying High Density Clusters
M. Anthony Wong · DSpace@MIT (Massachusetts Institute of Technology) · 2011
High density clusters are defined on a population with density f to be the maximal connected sets of values x with f(x)> c, for various values of c. It is desired to discover the high density clusters giver, a random sample of size N from the population. Using this clustering model, there is a correspondence between clustering and density estimation techniques. A hybrid algorithm is proposed which combines elements of both the k-means and single linkage techniques. This procedure is practicable for very large number of observations, and is shown to be consistent, under certain regularity conditions, in one dimension.