Enhanced SHAP Analysis for Multi-Class Classification Models

The SHAP plot for multi-class classification models is a visualization tool that uses the Shapley value to compute feature contributions for single predictions across multiple classes.

Usage

eSHAP_plot_multiclass(
  task,
  trained_model,
  splits,
  sample.size = 30,
  seed = 246,
  subset = 1
)

Arguments

task

mlr3 task object for multi-class classification

trained_model

mlr3 trained learner object

splits

mlr3 object defining data splits for train and test sets

sample.size

numeric, default to 30. The larger the value, the slower but more accurate the estimate of SHAP values

seed

numeric, an integer for reproducibility. Default to 246

subset

numeric, what percentage of the instances to use from 0 to 1 where 1 means all

Value

A list containing:

combined_plots

SHAP plot depicting the SHAP values for each class

shap_data

A matrix of SHAP values for each class.

combined_all_classes

overall SHAP plot depicting the SHAP values for all classes on a single plot

See also

Other classification: eSHAP_plot()

Other SHAP: eSHAP_plot()

Examples

# \donttest{
library("explainer")
seed <- 246
set.seed(seed)
# Load necessary packages and data...
# }