Weka set default cutoff (i.e. 50%) to discriminate its data of two class and can easily classify on different cutoff by using SelectThreshold classifier in Weka interface. Now in a three class data, Weka choose the highest probability, say from probability of 0.333, 0.421, 0.246, Weka will select 0.421. But in this class it is not possible to select a custom cutoff manually. Is there any method by which i can perform this task or some one can help me understanding the concept custom cutoff in multi-class data.
example of two class data
inst#, actual, predicted, error, probability distribution
1 1:+1 1:+1 *0.814 0.186
2 1:+1 1:+1 *1 0
3 1:+1 1:+1 *0.58 0.42
4 1:+1 2:-1 + 0.294 *0.706
5 1:+1 1:+1 *0.796 0.204
6 1:+1 1:+1 *0.989 0.011
7 1:+1 2:-1 + 0.294 *0.706
8 1:+1 1:+1 *0.814 0.186
example of three class data
inst#, actual, predicted, error, probability distribution
1 1:+1 1:+1 *0.92 0.07 0.01
2 1:+1 1:+1 *0.97 0.03 0
3 1:+1 2:0 + 0.47 *0.52 0.01
4 1:+1 1:+1 *0.89 0.08 0.03
5 1:+1 1:+1 *0.643 0.338 0.02
6 1:+1 2:0 + 0.22 *0.73 0.05
7 1:+1 1:+1 *0.45 0.21 0.34
8 1:+1 3:-1 + 0.16 0.19 *0.65
