fairmodels (version 1.2.1)

resample: Resample

Description

Method of bias mitigation. Similarly to reweight this method computes desired number of observations if the protected variable is independent from y and on this basis decides if this subgroup with certain class (+ or -) should be more or less numerous. Than performs oversampling or undersampling depending on the case. If type of sampling is set to 'preferential' and probs are provided than instead of uniform sampling preferential sampling will be performed. Preferential sampling depending on the case will sample observations close to border or far from border.

Usage

resample(protected, y, type = "uniform", probs = NULL, cutoff = 0.5)

Value

numeric vector of indexes

Arguments

protected

factor, protected variables with subgroups as levels (sensitive attributes)

y

numeric, vector with classes 0 and 1, where 1 means favorable class.

type

character, either (default) 'uniform' or 'preferential'

probs

numeric, vector with probabilities for preferential sampling

cutoff

numeric, threshold for probabilities

References

This method was implemented based on Kamiran, Calders 2011 https://link.springer.com/content/pdf/10.1007/s10115-011-0463-8.pdf

Examples

data("german")

data <- german

data$Age <- as.factor(ifelse(data$Age <= 25, "young", "old"))
y_numeric <- as.numeric(data$Risk) - 1

rf <- ranger::ranger(Risk ~ .,
  data = data,
  probability = TRUE,
  num.trees = 50,
  num.threads = 1,
  seed = 123
)

u_indexes <- resample(data$Age, y = y_numeric)

rf_u <- ranger::ranger(Risk ~ .,
  data = data[u_indexes, ],
  probability = TRUE,
  num.trees = 50,
  num.threads = 1,
  seed = 123
)

explainer_rf <- DALEX::explain(rf,
  data = data[, -1],
  y = y_numeric,
  label = "not_sampled"
)

explainer_rf_u <- DALEX::explain(rf_u, data = data[, -1], y = y_numeric, label = "sampled_uniform")

fobject <- fairness_check(explainer_rf, explainer_rf_u,
  protected = data$Age,
  privileged = "old"
)

fobject
plot(fobject)
# \donttest{
p_indexes <- resample(data$Age, y = y_numeric, type = "preferential", probs = explainer_rf$y_hat)
rf_p <- ranger::ranger(Risk ~ .,
  data = data[p_indexes, ],
  probability = TRUE,
  num.trees = 50,
  num.threads = 1,
  seed = 123
)

explainer_rf_p <- DALEX::explain(rf_p,
  data = data[, -1], y = y_numeric,
  label = "sampled_preferential"
)

fobject <- fairness_check(explainer_rf, explainer_rf_u, explainer_rf_p,
  protected = data$Age,
  privileged = "old"
)

fobject
plot(fobject)
# }