Some articles say that naive Bayes is naive because of "independence of attributes". Whereas others say "independence of attributes within a class". Can anybody please clear this confusion? Thanks
Naive Bayes doesn't assume independence of attributes ... It assumes conditional independence (or what you call independence within a class). This allows us to write the likelihood in the bayes rule P(X | Y) as the product of all P(Xi | Y), where X = (X1, ... Xi, ... Xn) and n is the number of attributes.