Summary

Using the Optimal Binning procedure, we have generated binning rules for scale variables that are potential predictors for the probability of default and applied these rules to a separate dataset.

During the binning process, you noted that the binned Years with current employer and Years at current address seem to do a better job of identifying high-probability non-defaulters, while the Credit card debt in thousands does a better job of identifying high-probability defaulters. This interesting observation will give you some extra insight when building predictive models for the probability of default. If avoiding bad debt is a primary concern, then Credit card debt in thousands will be more important than Years with current employer and Years at current address. If growing your customer base is the priority, then Years with current employer and Years at current address will be more important.