Playing With Probability Cut-off

Let’s consider that we are satisfied with our model after tuning it with the grid search module. The question is, Should we stay with the probability cutoff of 0.5 for the class prediction?

Overview

Sometimes, we might be interested in different probability cutoffs for the class predictions, typically in medical-related studies. We want to bring the high-risk patients for screening at a very early stage to avoid complications and reduce the long-term cost of treatments. Let’s look at the prediction from our selected model grid_search_rfc_pred once again.