fastshap: Error in UseMethod("explain") : no applicable method for 'explain' applied to an object of class "xgb.Booster"

Question

After fitting a xgboost model (model_n) I try to run the code below to obtain shap-values, where trainval is a dataframe with my traindata without the Y variabele:

shap_values <- explain(model_n, X = trainval)

I receive this error:

[[Error in UseMethod("explain") : 
  no applicable method for 'explain' applied to an object of class ["xgb.Booster"]]

If I use this line:

shap_values <- fastshap::explain(model_n, X = trainval)

I receive this error:

Error in explain.default(object, feature_names = feature_names, X = X,  : 
  argument "pred_wrapper" is missing, with no default

I have also tried with trainval as a matrix although this gives the same error.

Any idea what could be the cause?

Thanks a lot!

Answer 1

Adding library(vip) besides library(fastshap) to my code did solve the issue for using shap_values <- fastshap::explain(model_n, X = trainval)

Answer 2

fastshap is a good tool, but not a super-easy one. You need to pass a prediction function called pred_wrap so the fastshap explainer knows how to generate new predictions from your model.

Example code taken from https://bgreenwell.github.io/fastshap/articles/fastshap.html :

# Prediction wrapper
pfun <- function(object, newdata) {
  predict(object, data = newdata)$predictions
}

# Compute fast (approximate) Shapley values using 10 Monte Carlo repetitions
system.time({  # estimate run time
  set.seed(5038)
  shap <- explain(rfo, X = X, pred_wrapper = pfun, nsim = 10)
})

Also, if you are working with xgboost specifically the SHAPforxgboost package may be friendlier: https://liuyanguu.github.io/post/2019/07/18/visualization-of-shap-for-xgboost/

Finally the vip package is not a replacement for fastshap. It's a framework which contains many variable importance metrics and plots for them. Including, but not limited to shapley.