CLUSTERING USING GRAPH THEORY TOOLS

The aim of this paper is to introduce and describe a method for the classification of objects into a predetermined number of groups using the discrete mathematical tool graph theory. This clustering method is based on the most commonly used Euclidean distance, from which a complete graph is then constructed. From this complete graph, the minimum spanning tree of the graph is estimated. We have chosen this method because it is known that the minimum spanning tree can find certain structures in the data and therefore we assume that it also has a classification potential. Through this spanning tree, we seek to assign objects to groups that minimize the overall distance in the graph. The results of the assignment are then compared with the available hierarchical clustering methods, which are also based on the Euclidean distance for correctness. We present the results obtained by applying the method to the Iris dataset.