Reduce False positive while maintaining the size of prediction

Question

I\'m struggling with reducing false positive rate in my project. my project is basically an automated underwriting system that aims to accept more people automatically. Here