Problem with prediction using Feature Selection and Cross Validation


I am trying to build prediction system using Linear Regression combined with feature selection and cross validation. My data set consists of seven feature columns and one target column represented by decimal value (double).
My goal is to predict value of target column using aforementioned features.

Until now I have managed to build successful system using cross validation and Linear Regression learner-predictor model. Unfortunately when I am adding backward feature selection results are broken.
Scorer node produces confusion matrix full of “0” as both X and Y axis labels contains different decimal values which doesn’t match:

Final filter shows accuracy equal tu “0” at the output no matter which feature combination is used - my model looks like below and:

Is it possible to perform feature selection in data prediction task with this type of target values?
Thanks in advance for help!

Hello mackoks,

the scorer node is not suitable for a regression task such as yours.
You need to use the numeric scorer node.



Thank You very much Sir!
Works brilliant now. :slight_smile:

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