Prediction of Drought Severity Using Model-Based Clustering

Table 3

The performance of the MBCCDSS is detected by using initial state probabilities and without initial state probabilities.

Methods		= 1	= 2	= 3	= 4
EM algorithm (with )	LogL	−3539.739	−3467.271	−3447.74	−3435.261
	BIC	7142.191	7061.757	7087.199	7126.744
EM algorithm (without )	LogL	−3550.46	−3477.991	−3458.461	−3445.982
	BIC	7154.673	7065.28	7081.764	7112.35

The corresponding maximized LogL and BIC values are given for each number of components . BIC value provided in bold font shows the best model detected, which is compared to .