plot_explain_scatter: Visualize the explanation of interest (eoi) and LIME...

In goodekat/limeaid: Diagnose LIME Explanations

Description

Visualize the explanation of interest (eoi) and LIME simulated data

Usage

plot_explain_scatter(
  explanation,
  bins = TRUE,
  weights = TRUE,
  alpha = 1,
  line_size = 1,
  title.opt = TRUE
)

Arguments

explanation	Rows from the LIME explanations associated with one prediction of interest
bins	Should lines indicating the bins used by LIME be included? Only applicable is sim_method is equal to "quantile_bins" or "equal_bins". (Default is TRUE.)
weights	Should the size of the points represent the weight assigned by LIME? (Default is TRUE.)
alpha	Value to use for alpha blending of the points
line_size	Size of the lines used for representing the explainer model
title.opt	Should a title be included that lists the simulation method and Gower exponent? (Default is TRUE.)

Examples

# Prepare training and testing data
x_train = sine_data_train[c("x1", "x2", "x3")]
y_train = factor(sine_data_train$y)
x_test = sine_data_test[1:5, c("x1", "x2", "x3")]

# Fit a random forest model
rf <- randomForest::randomForest(x = x_train, y = y_train)
                       
# Run apply_lime
res <- apply_lime(train = x_train, 
                  test = x_test, 
                  model = rf,
                  label = "1",
                  n_features = 2,
                  sim_method = c('quantile_bins',
                                 'kernel_density'),
                  nbins = 2:3, 
                  return_perms = TRUE)
 
# Extract the rows associtaed with an explanation 
# of interest (the first observation in the test data) 
eoi <- res$explain[1:2,]

# Plot the explanation of interest
plot_explain_scatter(eoi)

goodekat/limeaid documentation built on March 26, 2021, 10:45 p.m.