Predicting breast cancer survivability: a comparison of three data mining methods