Forward Feature selection with Random forest classifier

[Baran]

Hi everyone,

I am using forward feature selection with Random forest classifier to predict that what features in my data is related to the class (Area-clusters) that I have. All the columns except “condition” and “cell names” are the feature that I want to test. I can’t get the configures right and even though the workflow works perfectly until the feature selection loop end, where it gives me the features with the score for accuracy, the feature selection filter just returns one feature. I uploaded the input table and the workflow. Any help and suggestion are much appreciated.
fff.xlsx (968.1 KB)
Cheers,

BaranForward Feature selection with Random forest classifier 3.knwf (22.1 KB)

[SimonS]

Hi @Baran,

that’s because having only feature ‘Col15’ included leads already to an accuracy of 1.0. Adding further features wouldn’t improve the model performance. Considering that you have 183 features, there might be something wrong with your data. You should check what ‘Col15’ is. It might be derived from your target column and should not be used for training. If not, congratulations! You reduced your feature space from 183 to 1

Cheers,
Simon