fastml (version 0.7.7)

explain_lime: Generate LIME explanations for a fastml model

Description

Creates a `lime` explainer using processed (encoded, scaled) data and returns feature explanations for new observations. The new observation is automatically preprocessed using the same recipe to ensure alignment with the explainer background.

Usage

explain_lime(
  object,
  new_observation,
  data = c("train", "test"),
  n_features = 5,
  n_labels = 1,
  ...
)

Value

An object produced by `lime::explain`.

Arguments

object

A `fastml` object.

new_observation

A data frame containing the new observation(s) to explain. Must contain the same columns as the original training data (before preprocessing). The function will apply the stored preprocessor to transform it.

data

Character string specifying which data to use for the LIME explainer background: "train" (default) or "test".

n_features

Number of features to show in the explanation. Default 5.

n_labels

Number of labels to explain (classification only). Default 1.

...

Additional arguments passed to `lime::explain`.

Examples

if (FALSE) {
data(iris)
iris <- iris[iris$Species != "setosa", ]
iris$Species <- factor(iris$Species)
model <- fastml(data = iris, label = "Species")
explain_lime(model, new_observation = iris[1, ])
}