Move the flag threshold and the risk metrics recompute from the model scores. Sample: the first 200 accounts of ar_default_data.csv.
Default rate by payment status
Pick the flag threshold
Flagged accounts
Precision, the share of flagged accounts that defaulted
Recall, the share of actual defaults that get flagged
Expected cost (units)
Findings
Confusion counts at the current threshold
Score one account
Type an account's predictor values and score it through the same coefficients. Check it: Account B (1.4, 0.6, 1, 0, 0.1, 4) should show P(default) = 0.3735.