You are developing a model to detect fraudulent credit card transactions. You need to prioritize detection, because missing even one fraudulent transaction could severely impact the credit card…

Increasing the score threshold would make the model more conservative, leading to more missed fraudulent transactions (false negatives), which is the opposite of the desired outcome.

Decreasing the score threshold makes the model more sensitive to the positive class, allowing predictions with lower confidence scores to be classified as fraudulent, thereby reducing false negatives.

Adding more positive examples to the training set helps the model learn the patterns associated with fraudulent transactions more effectively, especially in imbalanced datasets, which directly leads to improved detection of fraud.

Adding more negative (legitimate) examples would further imbalance the dataset if fraud is rare, potentially making it harder for the model to learn to identify fraudulent patterns.

Reducing the maximum number of node hours for training would prematurely stop the optimization process, likely decreasing model performance and worsening detection capabilities rather than improving them.