The present invention is a method for clustering data points. The method represents data-points as vertices of a graph (a well-known mathematical construct) with distance-weighted arcs (lines joining each paid of points). The method then involves sorting the arcs in increasing order of their weights and adding them in ascending order, at each stage determining the number of connected components in the graph and the length of the longest added edge. The longest edge is a measure of the quality of the clustering (low values are good), and the connected components are the clusters.