Minimal spanning tree based clustering technique: Relationship with Bayes Classifier

摘要

A minimal spanning tree (MST) based clustering technique along with its theoretical formulation is presented in this paper. The proposed technique is compared with Bayes Classifier and it is shown theoretically that the clustering technique, although an unsupervised one, approaches the performance of Bayes Classifier under a condition, as the number of sample points from each class increases. Experimental results with many synthetic data sets in 2-D and 3-D validate the theoretical prediction.