A data scientist trained an XGBoost model to flag fraudulent financial transactions. The training set is highly imbalanced (100,000 non-fraudulent vs. 1,000 fraudulent). On validation the model shows 99.1% accuracy but too many false negatives (fraudulent transactions predicted as non-fraudulent). Which two actions should the data scientist take to reduce false negatives?

The problem describes a highly imbalanced dataset where the model has high accuracy but too many false negatives, indicating it struggles to identify the minority class (fraud). Increasing scaleposweight in XGBoost directly addresses class imbalance by assigning a higher weight to the positive (fraudulent) class during training. This makes the model penalize misclassifications of the minority class more heavily, thereby reducing false negatives. Changing evalmetric to AUC (Area Under the ROC Curve) is also effective. AUC is a robust metric for imbalanced datasets because it evaluates the model's ability to distinguish between classes across various threshold settings, rather than relying on a single threshold like accuracy. Optimizing for AUC encourages the model to improve its true positive rate without significantly increasing the false positive rate, which helps reduce false negatives. Optimizing for RMSE is incorrect because RMSE is a regression metric, not suitable for binary classification. Increasing or decreasing maxdepth primarily addresses overfitting or underfitting, respectively, but doesn't directly tackle the class imbalance or the specific problem of high false negatives as effectively as scaleposweight or evalmetric for imbalanced classification.