Please use the source code attached to the previous bug report and refer to the second unit test in RealValueModelTests. The test passes what was tested for. But if you look at the results that are printed to standard out, they seem surprising to me. The first set of results that I get are:
classifiy with realModel: A = 0.586860
classifiy with repeatModel: A = 0.586860
classifiy with realModel: B = 0.413140
classifiy with repeatModel: B = 0.413140
This is strange to me because I have passed in "feature5" which is not even found in the training data for outcome B. I would expect the outcomes to be much more distinct (as in the following).
The second set of results that I get are:
classifiy with realModel: A = 0.998804
classifiy with repeatModel: A = 0.998804
classifiy with realModel: B = 0.001196
classifiy with repeatModel: B = 0.001196
The extreme difference in classification scores is surprising because I pass in all five features which are fairly well balanced between the two outcomes except for feature1 and feature5 which appear in only A and B, respectively.
Attached is a script that converts a real-valued data file to a repeated-value data file.