I have different results of "compare test results" in built model and separate "test" node.
I built regression model (split for test default 40%, no case_id).
The same entry data is connected to sample node (sample set to 40%, sample type random) and then to test node.
In test node I have predictive confidence 23% for SVN and 32% for GLM.
In regression model when I pick "compare test results" I have 9% from SVM and 10% for GLM.
Why these numbers are different ??
The Build node for Classification includes a internal stratified sample split.
If the data provided to the Test node is different from the internal split created by the Build node, which it sounds like it is in this case, then that would explain the difference.