K-means is a widely used method in cluster analysis, but what are its underlying assumptions and drawbacks? We examine what happens for non-spherical data and unevenly sized clusters.
