1
votes

I have a training set. I performed the following filters on the training set,

  1. Interquartile Range & RemoveWithValues(for removing outliers and extreme values).
  2. Replace Missing Values.
  3. NumerictoNominal

I used Naive Bayes classifier with cross validation(10 folds). Got an accuracy of 62%. Saved the model.

And I used the "NumerictoNominal" filter on test data. The test data has missing values and the Class attribute is also blank. The number and order of the attributes are the same in the test and training data.

And when I run Naive bayes using Supplied test set option in Weka I get an error saying training and test data are not compatible.

1

1 Answers

3
votes

See my answer to this question. You need same header information in test and training files.