Decision Tree: "Table contains no nominal attribute for classification"

New KNIME user here so apologies in advanced for what is probably a very simple new user problem.

I am trying to feed part of the Titanic data set into the decision tree node, but I get the error in the subject line above. Trying to use the random forest node gets me a similar error: "No possible target in input (node not connected?) -- unable to configure."

I've reduced the data set to just numeric values, converting sex and port of embarkation to ints and imputing the missing age values. I know I need to specify 'Survived' as the target but cannot open the configuration dialogs to do so.


- Mark

Hi Mark,

depending on the specific learner node you are using, it is ok to use a mix of nominal and numerical features as input, no need to convert them to numerical beforehand. In general a classification node will require the target variable (the one that you want to predict) to be nominal, while a regression node will require it to be numerical.

If you converted your target variable to numerical (int for example) and you are using a classification node with your data set, it is likely that the node will give a configuration error because it cannot find any nominal variable to work with. Check if this is the case.

Feel free to export your workflow and post it here if you would like to have a more precise diagnosis if what your problem could be.


That was the problem. The target variable in the original training data set is an integer (0 or 1 for did not or did survive). So the integer target column should be treated as nominal. I updated my file import to read that column as a string instead, and the model runs now.