1) Applying MultinomialNaivesBayes(not by any other classifier) in weka raises exception "problem evaluating classifier: Numeric attribute values must all be greater or equal to zero" ? How to fix it ?
2) Is dimensionality reduction (PCA, LSI, Random projection) is an alternative for feature selection (InformationGain, ChiSqr) or we need to apply both ? I have seen conflicting opinions about them on internet ?