Monotonic Binning for Credit Rating Models

Performs monotonic binning of numeric risk factor in credit rating models (PD, LGD, EAD) development. All functions handle both binary and continuous target variable. Functions that use isotonic regression in the first stage of binning process have an additional feature for correction of minimum percentage of observations and minimum target rate per bin. Additionally, monotonic trend can be identified based on raw data or, if known in advance, forced by functions' argument. Missing values and other possible special values are treated separately from so-called complete cases.

monobin 0.2.4

The goal of the monobin R package is to perform monotonic binning of numeric risk factor in credit rating models (PD, LGD, EAD) development. All functions handle both binary and continuous target variable. Missing values and other possible special values are treated separately from so-called complete cases.

Installation

You can install the released version of monobin from the CRAN executing the following code in R session:

install.packages("monobin")

In order to install latest version from github, you can use the following code:

library(devtools)
install_github("andrija-djurovic/monobin")

Example

This is a basic example which shows you how to solve a problem of monotonic binning of numeric risk factors:

suppressMessages(library(monobin))
data(gcd)
amount.bin <- cum.bin(x = gcd$amount, y = gcd$qual)
amount.bin[[1]]
gcd$amount.bin <- amount.bin[[2]]
gcd %>% group_by(amount.bin) %>% summarise(n = n(), y.avg = mean(qual))
#increase default number of groups (g = 20)
amount.bin.1 <- cum.bin(x = gcd$amount, y = gcd$qual, g = 20)
amount.bin.1[[1]]
#force trend to decreasing
cum.bin(x = gcd$amount, y = gcd$qual, g = 20, force.trend = "d")[[1]]

For more examples and package functions check the help page:

help(package = monobin)