Uncertainty quantification in classification problems: A Bayesian approach for predicating the effects of further test sampling