OptimalBinningWoE (version 1.0.8)

summary.obwoe: Summary Method for obwoe Objects

Description

Generates comprehensive summary statistics for optimal binning results, including predictive power classification based on established IV thresholds (Siddiqi, 2006), aggregate metrics, and feature-level diagnostics.

Usage

# S3 method for obwoe
summary(object, sort_by = "iv", decreasing = TRUE, ...)

Value

An S3 object of class "summary.obwoe" containing:

feature_summary

Data frame with per-feature statistics including IV classification (Unpredictive/Weak/Medium/Strong/Suspicious)

aggregate

Named list of aggregate statistics:

n_features

Total features processed

n_successful

Features without errors

n_errors

Features with errors

total_iv_sum

Sum of all feature IVs

mean_iv

Mean IV across features

median_iv

Median IV across features

mean_bins

Mean number of bins

iv_range

Min and max IV values

iv_distribution

Table of IV classification counts

target

Target column name

target_type

Target type (binary/multinomial)

Arguments

object

An object of class "obwoe".

sort_by

Character string specifying the column to sort by. Options: "iv" (default), "n_bins", "feature".

decreasing

Logical. Sort in decreasing order? Default is TRUE for IV, FALSE for feature names.

...

Additional arguments (currently ignored).

Details

IV Classification Thresholds

Following Siddiqi (2006), features are classified by predictive power:

Classification	IV Range
Unpredictive	< 0.02
Weak	0.02 - 0.10
Medium	0.10 - 0.30
Strong	0.30 - 0.50
Suspicious	> 0.50

Features with IV > 0.50 should be examined for data leakage or overfitting, as such high values are rarely observed in practice.

References

Siddiqi, N. (2006). Credit Risk Scorecards: Developing and Implementing Intelligent Credit Scoring. John Wiley & Sons. tools:::Rd_expr_doi("10.1002/9781119201731")

See Also

obwoe for the main binning function, print.obwoe, plot.obwoe.

Examples

# \donttest{
set.seed(42)
df <- data.frame(
  x1 = rnorm(500), x2 = rnorm(500), x3 = rnorm(500),
  target = rbinom(500, 1, 0.2)
)
result <- obwoe(df, target = "target")
summary(result)
# }