Resample Filter WEKA - How to interpret the result

Question

I am currently struggling with the problem of machine learning, while I have to deal with large, unbalanced datasets. That is, there are six classes ('1', '2' ... '6'). Unfortunately, for example, there are 150 examples / instances for classes '1', 90 for instances "2", and only 20 for class "3". All other classes cannot be "trained" because there are no instances available for these classes.

So far, I realized that WEKA (the machine learning tools I use) provides this controlled Resample filter. When I apply this filter with "noReplacement" = false and "bialToUniformClass" = 1.0, then this leads to a dataset where the number of instances is beautiful and almost equal (for class "1" .. "3" and others, stay empty).

Now my question is: how do WEKA and this filter generate "new" / additional instances for different classes.

Thank you for any hints or suggestions.

Cheers Julian

Answer 1

WEKA Resample . , .

, .

Answer 2

. . 2 1,0, N N , .

Answer 3

SMOTE .

, .