Research and application of clustering algorithm for arbitrary data set

This paper discusses the theory and algorithmic design of the CADD (clustering algorithm based on object density and direction) algorithm. This algorithm seeks to harness the respective advantages of the k-means and DENCLUE algorithms. Clustering results are illustrated using both a simple data set and one from the geological domain. Results indicate that CADD is robust in that automatically determines the number K of clusters, and is capable of identifying clusters of multiple shapes and sizes.