Mclust(x,G,···)

stance, G=2:5 indicates the Mclust function will construct four

models employing two, three, four and five clusters separately.

our models, the Mclust function will select the one which has

st BIC to return.

clust function was applied to a toy data with nine clusters data

igure 2.30(a). In total, 14 cluster models were tried by varying

er number from two to 15. The BIC values for these 14 models

ualised using the plot function as shown in Figure 2.36, where

seen that the model with nine clusters was the best.

The BIC values for using the mixture model algorithm to cluster a toy data with

rs shown in Figure 2.30(a).

e other clustering algorithms

also other clustering algorithms which have been developed and

o different areas. The K-medoids algorithm will estimate a

for each cluster of a cluster model [Kaufman and Rousseeuw,

iven N data points, ࣞൌሺܠଵ, ܠଶ, ⋯, ܠேሻ, a medoid of a cluster is

s below,

ܠ୫ୣୢ୭୧ୢൌargmin

ܡ∈ࣞ

෍݀ሺܡ, ܠ୧ሻ

ே

௜ୀଵ

(2.29)