Optimizing the Computation of Approximate Certain Query Answers over Incomplete Databases