ws the F statistic p value distribution for using the fuzzy C means

m based on different cluster numbers for the data set with nine

which is shown in Figure 2.30(a). In total, 15 models were tested.

models employed two, three, till 16 clusters. Based on the F

p value distribution, the model with eight clusters was also

as the best model structure, which was again one cluster less than

luster number.

The F statistic p value distribution of the fuzzy C-means models constructed for

with nine clusters, where 15 models with employed two, three, till 16 clusters

.

e 2.32 shows a comparison between a K-means model and a fuzzy

model applied to a data set with three clusters. It can be seen that

se, two models show a similar performance.

(a) (b)

he comparison between a K-means model and a fuzzy C-means model for a toy

hree clusters. Three classes of data were labelled by A, B and C. The circles, the

nd the diamonds were used to label predicted clusters. (a) The K-means model.

zy C-means model.