Figure 6

Figure of Merit. Figure of Merit (FOM) has been delineated by Yeung et al. [46] particularly for the analysis of gene expression data. The index estimates the predictive power of a cluster algorithm, and thereby revealed HCA using Ward's linkage and Euclidean distances, K-means as well as Neuralgas as the best performing cluster algorithms, while correlation-based cluster algorithms, Single-Linkage using Euclidean distances, and--at least in two experiments--fuzzy C-means produce the least reliable results. The lowest observed value is marked with an asterisk symbol, whereas at the position of the highest index value the corresponding cluster number is printed.