Write a Scorecard Workbook

Turns a fitted obwoe_scorecard into a multi-sheet .xlsx file: the points table a branch officer reads, the evidence a validation team asks for, and the SQL a deployment engineer runs.

Usage

obwoe_report(
  x,
  file,
  control = NULL,
  table = "your_table",
  dialect = "ansi",
  keep_columns = NULL
)

Arguments

x

An "obwoe_scorecard" object.

file

Path to the .xlsx file to write.

control

An "obwoe_scorecard_control" object; only digits and overwrite are used. Defaults to the settings stored on x where available.

table

Character string naming the source table used in the generated SQL. Default "your_table".

dialect

SQL dialect for the deployment sheets, passed to obwoe_sql. Default "ansi".

keep_columns

Columns the generated SQL carries through unchanged.

Value

The path written, invisibly.

Details

The workbook has one sheet per stage, in the order a reviewer reads them:

01_Model_Summary	provenance, the funnel, headline metrics
02_Scorecard	the deliverable: one row per variable and bin, with the integer points
03_Coefficients	the fit, with the sign check and standard errors
04_Bin_Statistics	the gains table of every binned variable used in training
05_Screening	every candidate and why it lived or died
06_Correlations	redundancy in the WoE space and what was pruned
07_Score_Gains	rank ordering per sample, with observed versus predicted
08_Stability_PSI	the score and every variable, per sample
09_Cutoff_Strategy	approval rate, bad rate and swap set by cutoff
10_SQL_WoE	obwoe_sql output: the WoE transform
11_SQL_Points	the same bins returning integer points, summed into the score
12_Reproducibility	the call, versions, and every warning raised

Splitting the deployment SQL in two is deliberate. The WoE sheet reproduces the model exactly and is what a data scientist re-scores with; the points sheet reproduces the card, which is what the business signed and what the decision engine should run.

See also

obwoe_scorecard, obwoe_sql

Examples

# \donttest{
german <- read.csv(
  gzfile(system.file("extdata", "germancredit.csv.gz",
    package = "OptimalBinningWoE"
  )),
  stringsAsFactors = FALSE
)
german$default <- 1L - german$credit_risk
german$credit_risk <- NULL

sc <- obwoe_scorecard(german, target = "default", seed = 1)
#> Warning: 210 events for 13 variables (16.2 per variable): below the 20-events-per-variable rule of thumb, so the coefficients are unstable.
obwoe_report(sc, file = file.path(tempdir(), "scorecard.xlsx"))
# }