scorecard

CRAN_Status_Badge Travis build status

The goal of scorecard package is to make the development of traditional credit risk scorecard model easier and efficient by providing functions for some common tasks. - data partition (split_df) - variable selection (iv, var_filter) - weight of evidence (woe) binning (woebin, woebin_plot, woebin_adj, woebin_ply) - scorecard scaling (scorecard, scorecard_ply) - performance evaluation (perf_eva, perf_psi)

Installation

install.packages("scorecard")
# install.packages("devtools")
devtools::install_github("shichenxie/scorecard")

Example

This is a basic example which shows you how to develop a common credit risk scorecard:

# Traditional Credit Scoring Using Logistic Regression
library(scorecard)

# data prepare ------
# load germancredit data
data("germancredit")

# filter variable via missing rate, iv, identical value rate
dt_s = var_filter(germancredit, y="creditability")

# breaking dt into train and test
dt_list = split_df(dt_s, y="creditability", ratio = 0.6, seed = 30)
train = dt_list$train; test = dt_list$test;

# woe binning ------
bins = woebin(dt_s, y="creditability")
# woebin_plot(bins)

# binning adjustment
# # adjust breaks interactively
# breaks_adj = woebin_adj(dt_s, "creditability", bins) 
# # or specify breaks manually
breaks_adj = list(
  age.in.years=c(26, 35, 40),
  other.debtors.or.guarantors=c("none", "co-applicant%,%guarantor"))
bins_adj = woebin(dt_s, y="creditability", breaks_list=breaks_adj)

# converting train and test into woe values
train_woe = woebin_ply(train, bins_adj)
test_woe = woebin_ply(test, bins_adj)

# glm ------
m1 = glm( creditability ~ ., family = binomial(), data = train_woe)
# summary(m1)

# # Adjusting for oversampling (support.sas.com/kb/22/601.html)
# library(data.table)
# p1=0.03; r1=0.3
# dt_woe = dt_woe[, weight := ifelse(y==1, p1/r1, (1-p1)/(1-r1) )]
# 
# fmla = as.formula(paste("y ~", paste(names(dt_woe)[2:21], collapse="+")))
# m1 = glm(fmla, family = binomial(), data = dt_woe, weights = weight)


# Select a formula-based model by AIC (or by LASSO)
m_step = step(m1, direction="both", trace = FALSE)
m2 = eval(m_step$call)
# summary(m2)

# performance ks & roc ------
# predicted proability
train_pred = predict(m2, train_woe, type='response')
test_pred = predict(m2, test_woe, type='response')
# performance
train_perf = perf_eva(train$creditability, train_pred, title = "train")
test_perf = perf_eva(test$creditability, test_pred, title = "test")

# score ------
card = scorecard(bins_adj, m2)
# credit score
train_score = scorecard_ply(train, card, print_step=0)
test_score = scorecard_ply(test, card, print_step=0)

# psi
perf_psi(
  score = list(train = train_score, test = test_score),
  label = list(train = train$creditability, test = test$creditability)
)