Fig. 1.

The main analysis workflow consisted of four stages. First, the clustering algorithms were applied to the eight cancer datasets to generate clustering partitions (blue). Then seven different metrics of clustering quality were examined and grouped into three distinct groups by similarity (yellow). By combining three representative measures, one per group, we generated quality scores first for each clustering partition and then for each algorithm (green). Finally, we ranked the algorithms by quality scores for each choice of measures, and then combined these ranks into a final ranking (pink). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)