Problem with Feature Selection and Classification

I am trying to complete a classification workflow which contains feature selection and category to number conversion. The problem I am facing is that I am using the Feature Selection Loop Start 1:1 node with the Random method, then I am doing partitioning for training and testing sets (here starts the problem when I want to select a feature for stratified sampling the complete list of feature). Then when I am using Category To Number only a subset of the features are there in the “include” section, and this subset changes according to the method for feature selection.
What I am expecting is the complete list of features to choose from and it was working before as shown in the first screenshot, while the other 2 screenshots illustrate the problem.
Screen Shot 2022-04-29 at 14.03.14


Hi,

Can you provide a workflow (knime file, .knar file)? Also provide fake or real data (or use a Table Creator node) as well so we can better assist you. Thanks

At first glance it sounds like when you used feature selection you were left with less features and therefore when you inspect those subsequent nodes, they only contain the selected features. Is that right?

BIA505_midtermProjectClassification.knwf (45.8 KB)

Could it be that there is a threshold for category to number node? So if there are too many distinct values in a column it does not work anymore and that is why the columns can’t be selected?
br

Can you supply the CSV you used?

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.