Scikit-learn Random Forest Classifier: combining numeric values with multi-labels

I have a training dataset where every feature contains two numeric values and six (out of a possible twelve) unique categorical values. What I want to do is train up a random forest classifier using the feature's two numeric values while assigning each feature six labels, in the aim that for my test values, I can figure out which of the labels most correlate with the numeric values.

Am I right in thinking that the 'forest.fit(feature[numeric data], feature[label data])' is the right approach? When I try and score my data, I get the following error:

ValueError: multiclass-multioutput is not supported

So I'm not putting my labels in correctly. Score(X,y) - X is my two numeric values as floats, my y array is a pandas dataframe containing the labels [1,2,3,7,8,9]