scorer does not shows the total number of instances

[Monira_Mahmoud]

02RandomForestClassification.knwf (17.4 KB)
the scorer not shows that all instances used in analysis. the data set includes 9104 instances and the scorer shows

could you help please?

[qqilihq]

Your downloadable workflow unfortunately contains no data. Just by looking at the structure: Keep in mind that you’re partitioning your data using the Partitioning node, thus the classifier only runs on a subset of your entire dataset.

–Philipp

[Monira_Mahmoud]

thanks a lot Philippi,
I am new to knime,get used to use WEKA. I got that the predictor works only in the test partition. but how could make the predictor shows the score on all the data set like in WEKA?
thanks again

[qqilihq]

Hi Monira,

@Monira_Mahmoud

but how could make the predictor shows the score on all the data set like in WEKA?

I don’t know how Weka does it, to be honest. You could of course run the entire data set through the predictor and then score it. But I’m quite sure that this is not what you’ll really want, as you’d be scoring the training data as well, which would give you overly optimistic results Scoring only the unseen data, is the right way to do it.

If you strictly want to score with all data, I suggest you look into “n-fold cross validation” where you train/test sequentially with several split and aggregate the evaluation results. KNIME seems to have some nodes for that too (never used them personally):

Cross Validation — NodePit

Hope this helps!

[Monira_Mahmoud]

WEKA do it through cross validation. Do you think that I need to use the cross validation node before or after partitioning?

[qqilihq]

Yes, just give it a try

[Monira_Mahmoud]

thank you will give it a go.

[Daniel_Weikert]

Hi I do not know Weka either but there is no need to partition if you do cross validation.It should do the splitting within the cross validation
br