Consider a data set with four binary attributes X1, X2, X3 and X4. The attribute X4 takes exactly the same value as X3 for each record, i.e., X4 is equal to X3.

We build two Naive Bayes models: one using all four attributes X1, X2, X3 and X4 and the other using only three attributes X1, X2 and X3.

a) Do you expect the predictions for a new test instance using the two Naive Bayes models to be same or different? Briefly explain.

b) Do you expect the naive Bayes classifier built using all 4 attributes to be better or worse than the naive Bayes classifier built using 3 attributes? Briefly explain.