Merge pull request #487 from tidymodels/try-catch-augment

`augment()` for models without class probabilities

Author: topepo

R/aaa_models.R

@@ -286,9 +286,13 @@ check_pred_info <- function(pred_obj, type) {
   invisible(NULL)
 }
 
-check_spec_pred_type <- function(object, type) {
+spec_has_pred_type <- function(object, type) {
   possible_preds <- names(object$spec$method$pred)
-  if (!any(possible_preds == type)) {
+  any(possible_preds == type)
+}
+check_spec_pred_type <- function(object, type) {
+  if (!spec_has_pred_type(object, type)) {
+    possible_preds <- names(object$spec$method$pred)
     rlang::abort(c(
       glue::glue("No {type} prediction method available for this model."),
       glue::glue("Value for `type` should be one of: ",

R/augment.R

@@ -6,8 +6,9 @@
 #' [fit()] and `new_data` contains the outcome column, a `.resid` column is
 #' also added.
 #'
-#' For classification models, the results include a column called `.pred_class`
-#' as well as class probability columns named `.pred_{level}`.
+#' For classification models, the results can include a column called
+#'  `.pred_class` as well as class probability columns named `.pred_{level}`.
+#'  This depends on what type of prediction types are available for the model.
 #' @param x A `model_fit` object produced by [fit()] or [fit_xy()].
 #' @param new_data A data frame or matrix.
 #' @param ... Not currently used.
@@ -56,6 +57,7 @@
 #'
 augment.model_fit <- function(x, new_data, ...) {
   if (x$spec$mode == "regression") {
+    check_spec_pred_type(x, "numeric")
     new_data <-
       new_data %>%
       dplyr::bind_cols(
@@ -68,12 +70,18 @@ augment.model_fit <- function(x, new_data, ...) {
       }
     }
   } else if (x$spec$mode == "classification") {
-    new_data <-
-      new_data %>%
-      dplyr::bind_cols(
-        predict(x, new_data = new_data, type = "class"),
+    if (spec_has_pred_type(x, "class")) {
+      new_data <- dplyr::bind_cols(
+        new_data,
+        predict(x, new_data = new_data, type = "class")
+      )
+    }
+    if (spec_has_pred_type(x, "prob")) {
+      new_data <- dplyr::bind_cols(
+        new_data,
         predict(x, new_data = new_data, type = "prob")
       )
+    }
   } else {
     rlang::abort(paste("Unknown mode:", x$spec$mode))
   }

man/augment.Rd

@@ -76,3 +76,18 @@ test_that('classification models', {
 
 })
 
+
+test_that('augment for model without class probabilities', {
+  skip_if_not_installed("LiblineaR")
+
+  data(two_class_dat, package = "modeldata")
+  x <- svm_linear(mode = "classification") %>% set_engine("LiblineaR")
+  cls_form <- x %>% fit(Class ~ ., data = two_class_dat)
+
+  expect_equal(
+    colnames(augment(cls_form, head(two_class_dat))),
+    c("A", "B", "Class", ".pred_class")
+  )
+  expect_equal(nrow(augment(cls_form, head(two_class_dat))), 6)
+
+})