The interesting part isn't the logistic regression — it's the bucketing. Naive binning ignores where risk actually changes. My DP implementation finds the boundaries that maximize log-likelihood of…