How to choose the proper metrics for False Positive Reduction

Hi, I am dealing with a false positive reduction problem, and the ratio of positive and negative is 1.7:1.

Firstly, I wonder what can I do to deal with slightly skewed training data? Though Applying weights to different classes, and over/undersampling of the data are already noted, shall I try those tricks independently? Or shall I do nothing at all?

At the moment I have difficulty finding the proper metrics to evaluate my models. Any materials like github repos, or kaggle kernel would be much appreciated.