perf_psi: PSI

Description

perf_psi calculates population stability index (PSI) for both total credit score and variables. It can also creates graphics to display score distribution and bad rate trends.

Usage

perf_psi(score, label = NULL, title = NULL, show_plot = TRUE,
  positive = "bad|1", threshold_variable = 20, var_skip = NULL, ...)

Arguments

score

A list of credit score for actual and expected data samples. For example, score = list(actual = scoreA, expect = scoreE).

label

A list of label value for actual and expected data samples. For example, label = list(actual = labelA, expect = labelE). Default is NULL.

title

Title of plot, default is NULL.

show_plot

Logical. Default is TRUE.

positive

Value of positive class, default is "bad|1".

threshold_variable

Integer. Default is 20. If the number of unique values > threshold_variable, the provided score will be counted as total credit score, otherwise, it is variable score.

var_skip

Name of variables that are not score, such as id column. It should be the same with the var_kp in scorecard_ply function. Default is NULL.

...

Additional parameters.

Value

A data frame of psi and graphics of credit score distribution

Details

The population stability index (PSI) formula is displayed below: $$PSI = \sum((Actual\% - Expected\%)*(\ln(\frac{Actual\%}{Expected\%}))).$$ The rule of thumb for the PSI is as follows: Less than 0.1 inference insignificant change, no action required; 0.1 - 0.25 inference some minor change, check other scorecard monitoring metrics; Greater than 0.25 inference major shift in population, need to delve deeper.

Examples

Run this code

# NOT RUN {
# data preparing ------
# load germancredit data
data("germancredit")
# filter variable via missing rate, iv, identical value rate
dt_f = var_filter(germancredit, "creditability")
# breaking dt into train and test
dt_list = split_df(dt_f, "creditability")
label_list = lapply(dt_list, function(x) x$creditability)

# woe binning ------
bins = woebin(dt_list$train, "creditability")
# converting train and test into woe values
dt_woe_list = lapply(dt_list, function(x) woebin_ply(x, bins))

# glm ------
m1 = glm(creditability ~ ., family = binomial(), data = dt_woe_list$train)
# vif(m1, merge_coef = TRUE)
# Select a formula-based model by AIC
m_step = step(m1, direction="both", trace=FALSE)
m2 = eval(m_step$call)
# vif(m2, merge_coef = TRUE)

# predicted proability
pred_list = lapply(dt_woe_list, function(x) predict(m2, type = 'response', x))

# scorecard ------
card = scorecard(bins, m2)

# credit score, only_total_score = TRUE
score_list = lapply(dt_list, function(x) scorecard_ply(x, card))
# credit score, only_total_score = FALSE
score_list2 = lapply(dt_list, function(x) scorecard_ply(x, card, only_total_score=FALSE))


###### perf_eva examples ######
# Example I, one datset
## predicted p1
perf_eva(pred = pred_list$train, label=dt_list$train$creditability, title = 'train')
## predicted score
# perf_eva(pred = score_list$train, label=dt_list$train$creditability, title = 'train')

# Example II, multiple datsets
## predicted p1
perf_eva(pred = pred_list, label = label_list)
## predicted score
# perf_eva(score_list, label_list)


###### perf_psi examples ######
# Example I # only total psi
psi1 = perf_psi(score = score_list, label = label_list)
psi1$psi  # psi data frame
psi1$pic  # pic of score distribution

# Example II # both total and variable psi
psi2 = perf_psi(score = score_list, label = label_list)
# psi2$psi  # psi data frame
# psi2$pic  # pic of score distribution


###### gains_table examples ######
# Example I, input score and label can be a list or a vector
gains_table(score = score_list$train, label = label_list$train)
gains_table(score = score_list, label = label_list)

# Example II, specify the bins number and type
gains_table(score = score_list, label = label_list, bin_num = 20)
gains_table(score = score_list, label = label_list, bin_type = 'width')
# }