The question marks in accuracy table

Hello everyone. I have a question regarding the question marks under the accuracy column. Could someone explain to me why the question marks show up even though there are no values that could cause undefined accuracy measure? All values in the table are able to be involved in a measure.

Thanks in advance.

Question marks corresponds to missing values or nulls from databases. To manage them use


node.

Hi,
In the Scorer node there simply is no accuracy for individual classes. Here you can check precision, recall, etc. instead. Since Accuracy is a column but only valid for the Overall row, all other rows contain missing values here. So this is no error and will always be like that, no matter the model.
Kind regards
Alexander

3 Likes

Oh, I thought it also represented the individual columns. Actually, I’m new in data mining practices. My intention is to predict what could the " not specified " records of passengers’ travel purpose be. I used the random forest model for the classification. But I guess this is not how I can predict a specific outcome (“not specified” since this model only divides the dataset as train and test data and then predicts the already labeled entries. I am confused about how to interpret my findings actually.
Here is my model


According to this confusion matrix
confusion matrix
I intended to interpret as "the model predicted 446 “not specified” labeled entries is predicted as “leisure”, 33 entries as “work”,… "
but it sounds quite odd to me, it feels like I’m mistaken.

Hi,
I think what you need to do first is separate your “not specified” rows from the others using the Row Splitter and train your model only on the cases where the purpose is given. This part of the data you should split again into Test and Training set and then for the training I would use cross validation to find a good model. Once you have one that scores well, train it on all your Training data and see how well it performs on the Test set. If it is good there, train it again on Test and Training set and use it to predict the “not specified” cases.
Kind regards
Alexander

3 Likes

Hello,
thank you for your help.

Best regards,
Selin

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.