Classification tools (e.g., Bayesian classifiers, neural networks, etc.) are usually unable to perfectly classify all problem instances even if large amounts of data are available for learning the models. As a result, firms often must consider the costs associated with incorrect predictions when using such models. Consider a binary classification problem (e.g., approve or deny loan applications) where the cost of incorrect classifications are the same for both classes. What would be the probability threshold based on which loan applications should be approved (or denied)? What would be the probability threshold for approving (or denying) loans if the cost of incorrectly approving an applicant is three times as much as the cost of incorrectly denying an applicant? What would be the probability threshold for approving (or denying) loans if the cost of incorrectly denying an applicant is three times as much as the cost of incorrectly approving an applicant?