R/shap.R

In jmaspons/MLTools: Pipes and tools for machine learning

get_shap <- function(model, test_data, bg_data, bg_weight=NULL, verbose=TRUE) {
  if (inherits(model, "randomForest") && model$type == "classification") {
    warning("SHAP not implemented for randomForest with classificacion trees.")
    return (NULL)
  } else {
    shap <- kernelshap::kernelshap(object = model, X = test_data, bg_X = bg_data, bg_w = bg_weight, verbose = verbose)
    sv <- shapviz::shapviz(shap)
  }

  return(sv)
}