model_parts function

Dataset Level Variable Importance as Change in Loss Function after Variable Permutations

From DALEX version 1.0 this function calls the feature_importance

Find information how to use this function here: https://ema.drwhy.ai/featureImportance.html.

model_parts(
  explainer,
  loss_function = loss_default(explainer$model_info$type),
  ...,
  type = "variable_importance",
  N = n_sample,
  n_sample = 1000
)

Arguments

• explainer: a model to be explained, preprocessed by the explain function
• loss_function: a function that will be used to assess variable importance. By default it is 1-AUC for classification, cross entropy for multilabel classification and RMSE for regression. Custom, user-made loss function should accept two obligatory parameters (observed, predicted), where observed states for actual values of the target, while predicted for predicted values. If attribute "loss_accuracy" is associated with function object, then it will be plotted as name of the loss function.
• ...: other parameters
• type: character, type of transformation that should be applied for dropout loss. variable_importance and raw results raw drop lossess, ratio returns drop_loss/drop_loss_full_model while difference returns drop_loss - drop_loss_full_model
• N: number of observations that should be sampled for calculation of variable importance. If NULL then variable importance will be calculated on whole dataset (no sampling).
• n_sample: alias for N held for backwards compatibility. number of observations that should be sampled for calculation of variable importance.

Returns

An object of the class feature_importance. It's a data frame with calculated average response.

Examples

# regression

library("ranger")
apartments_ranger_model <- ranger(m2.price~., data = apartments, num.trees = 50)
explainer_ranger  <- explain(apartments_ranger_model, data = apartments[,-1],
                             y = apartments$m2.price, label = "Ranger Apartments")
model_parts_ranger_aps <- model_parts(explainer_ranger, type = "raw")
head(model_parts_ranger_aps, 8)
plot(model_parts_ranger_aps)

# binary classification

titanic_glm_model <- glm(survived~., data = titanic_imputed, family = "binomial")
explainer_glm_titanic <- explain(titanic_glm_model, data = titanic_imputed[,-8],
                         y = titanic_imputed$survived)
logit <- function(x) exp(x)/(1+exp(x))
custom_loss <- function(observed, predicted){
   sum((observed - logit(predicted))^2)
}
attr(custom_loss, "loss_name") <- "Logit residuals"
model_parts_glm_titanic <- model_parts(explainer_glm_titanic, type = "raw",
                                       loss_function = custom_loss)
head(model_parts_glm_titanic, 8)
plot(model_parts_glm_titanic)

# multilabel classification

HR_ranger_model_HR <- ranger(status~., data = HR, num.trees = 50,
                               probability = TRUE)
explainer_ranger_HR  <- explain(HR_ranger_model_HR, data = HR[,-6],
                             y = HR$status, label = "Ranger HR")
model_parts_ranger_HR <- model_parts(explainer_ranger_HR, type = "raw")
head(model_parts_ranger_HR, 8)
plot(model_parts_ranger_HR)

References

Explanatory Model Analysis. Explore, Explain and Examine Predictive Models. https://ema.drwhy.ai/