A fast algorithm for robust constrained clustering

The application of “concentration” steps is the main principle behind Forgy’s kmeans algorithm and Rousseeuw and van Driessen’s fast-MCD algorithm. Although they share this principle, it is not completely straightforward to combine both algorithms for developing a clustering method which is not affected by a certain proportion of outlying observations and that is able to cope with non spherical groups or with groups with different weights. However, these approaches can be successfully combined by additionally controlling the relative cluster scatters in the concentration steps. In this way, the appearance of uninteresting spurious clusters is avoided. An algorithm which implements such “constrained concentration” steps in a computationally efficient way will be presented in this work.