cutoff.palette: Palette of cutoff values that minimize and maximize metrics...

In PDtoolkit: Collection of Tools for PD Rating Model Development and Validation

Palette of cutoff values that minimize and maximize metrics from the confusion matrix

Description

cutoff.palette returns confusion matrix along with accompanied performance metrics.

Usage

cutoff.palette(predictions, observed, min.pct.obs = 0.05, min.pct.def = 0.01)

Arguments

predictions	Model predictions.
observed	Observed values of target variable.
min.pct.obs	Minimum percentage of observations. Used to select boundaries of cutoff values. Default value is 0.05.
min.pct.def	Minimum percentage of default. Used to select boundaries of cutoff values. Default value is 0.01.

Value

The command cutoff.palette returns data frame with minimum and maximum values of each confusion matrix metric along with optimized cutoff itself.

Examples

suppressMessages(library(PDtoolkit))
data(loans)
#identify numeric risk factors
num.rf <- sapply(loans, is.numeric)
num.rf <- names(num.rf)[!names(num.rf)%in%"Creditability" & num.rf]
#discretized numeric risk factors using mdt.bin from monobin package
loans[, num.rf] <- sapply(num.rf, function(x) 
			  mdt.bin(x = loans[, x], y = loans[, "Creditability"])[[2]])
str(loans)
res <- stepFWD(start.model = Creditability ~ 1, 
   p.value = 0.05, 
   coding = "WoE",
   db = loans)
#run cutoff optimization
cop <- cutoff.palette(predictions = predict(res$model, type = "response"), 
			observed = loans$"Creditability",
			min.pct.obs = 0.05, 
			min.pct.def = 0.01)
cop
confusion.matrix(predictions = predict(res$model, type = "response"), 
	     observed = loans$"Creditability",
	     cutoff = cop$cutoff.max[cop$metric%in%"f1.score"])

PDtoolkit documentation built on Sept. 20, 2023, 9:06 a.m.