In this paper we propose a measure of clustering quality or accuracy that is
appropriate in situations where it is desirable to evaluate a clustering
algorithm by somehow comparing the clusters it produces with ``ground truth'
consisting of classes assigned to the patterns by manual me