Cross-validation testing is a statistical method used to assess how well a model generalizes to an independent data set. The AAIA™ Study Guide recommends this method as a best practice to fine-tune model accuracy and reduce both false positives and false negatives. It involves splitting the dataset into training and testing subsets multiple times to ensure model robustness.

“Cross-validation allows auditors and developers to identify overfitting and adjust model parameters to achieve better generalization and predictive accuracy, especially in fraud detection contexts.”

Regression testing (A) focuses on changes over time; substantive testing (C) is audit-specific but not model-focused. Benford’s Law (D) applies to numerical distributions but is not designed for optimizing ML models. Hence, B is the best approach.