Problem using Numeric Scorer with PNN Learner

Dear, KNIME community,
I face with following problem, I perform the same operations on same students data using PNN Learner and Numeric Scorer. The only difference: I am trying to predict two different characteristics NR_OF_FAILURES and WEIGHTED_AVERAGE. It works perfect for WEIGHTED_AVERAGE and did not work for NR_OF_FAILURES (WARN - Can’t calculate Mean Absolute Percentage error: target value is 0! Row1). What I am doing wrong?
TEST.xlsx (4.1 KB)
Student rusults.knwf (21.9 KB)

Hi @AlexVilnius -

I believe the problem here arises from the calculation of the MAPE. If you look in the Numeric Scorer node description, you can scroll down and see that in the denominator, you have your forecast value. Well, in this case, both of your forecast values are zero, so the MAPE is undefined. :slight_smile:

As it is, building a model based on only 6 observations is questionable anyway. So I would try inputting a larger dataset with non-zero values, and you most likely won’t have this problem.


This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.