What will happen if you choose low support and high confidence and vice-versa for generating rules by association rule mining for a dense dataset?