KAscore

The goal of KAscore is to help analysts at financial institutions develop, manage, and deploy credit scorecard models.

Background

This R package is heavily based on the book Intelligent Credit Scoring: Building and Implementing Better Credit Risk Scorecards by Naeem Siddiqi (2nd Edition).

There are other R and Python packages for building credit scorecards that lack documentation and testing and have opinionated logic. The goal of {KAscore} is to provide ample documentation around the package’s functions, rigorous testing to ensure validity of logic, flexibility in the type of scorecard that can be developed, and ease-of-use for end users.

Installation

You can install the package from source (.tar.gz file provided by Ketchbrook) by running:

# Install {remotes} package if not installed already
# install.packages("remotes")

# Install {KAscore}
remotes::install_local("local/path/to/KAscore_v1.2.0.tar.gz")

Using {renv}

If you use {renv} for project-based isolated R environments, this vignette provides a few different ways you can install a local R package (.tar.gz file) into your {renv} project.

Getting Started

This package contains a built-in mock dataset called loans to provide users with a quick way to try out the functionality of the package itself.

library(KAscore)

# Take a look at the first few rows of data
head(loans) |> 
  knitr::kable()

loan_id	amount_of_existing_debt	term	industry	loan_amount	other_debtors_guarantors	years_at_current_address	collateral_type	housing_status	count_loan_facilities	default_status
100100	[$250,000, Inf)	6	grain	11690	none	[15, Inf)	real estate	own	2	good
100101	[$50,000, $250,000)	48	grain	59510	none	[5, 10)	real estate	own	1	bad
100102	[$0, $10,000)	12	pork	20960	none	[10, 15)	real estate	own	1	good
100103	[$250,000, Inf)	42	dairy	78820	guarantor	[15, Inf)	building society savings agreement/ life insurance	rent	1	good
100104	[$250,000, Inf)	24	fruit	48700	none	[15, Inf)	unknown / no property	rent	2	bad
100105	[$0, $10,000)	36	pork	90550	none	[15, Inf)	unknown / no property	rent	1	good

The best way to get started with {KAscore} is to read the Articles, or reach out to Ketchbrook Analytics directly to set up a demo, at info@ketchbrookanalytics.com.