How to filter out one class of the label?

Miner_Markus

Hi together,

I want to analyze some data in regard to a label that has 3 classes. Due to bad data quality in one of the classes and because it is also not of interest, I want to focus only on 2 classes of the label. I used "Filter Example" to delete all examples, that are labeled with the label, that is not of interest. But in my model (decision tree) it is still displayed when I click on the leafs, for examples as (25/8/0) or (6/32/0) etc. I would like to eliminate the 0 in those brackets.

Somebody has an idea?

varunm1

Hello @Miner_Markus

Once you filter the examples, then connect the "exa" port of filter examples operator to another operator called "Guess Types". In the parameters of guess types operator select attribute filter type "Single" and in attribute select your desired label attribute. Also, you should tick "include special attributes". This will remove that class from your tree.

This trick from @IngoRM works well. I scratched my head so hard to remember this operator name.

Miner_Markus

Perfect, thank you!

MartinLiebig

Hi @varunm1 and @Miner_Markus ,

while this works, the usual way to do this is the Remove Unused Values operator. I think guess types calls this in the backend, but also does a few more things.

BR,

Martin