Github user jkbradley commented on a diff in the pull request:
https://github.com/apache/spark/pull/6339#discussion_r35806884
--- Diff:
mllib/src/main/scala/org/apache/spark/ml/feature/StringIndexer.scala ---
@@ -151,4 +153,77 @@ class StringIndexerModel private[ml] (
val copied = new StringIndexerModel(uid, labels)
copyValues(copied, extra)
}
+
+ /**
+ * Return a model to perform the inverse transformation.
+ * Note: By default we keep the original columns during this
transformation, so the inverse
+ * should only be used on new columns such as predicted labels.
+ */
+ def invert(inputCol: String, outputCol: String):
StringIndexerInverseTransformer = {
+ val labelsCol: String = $(this.outputCol)
+ new StringIndexerInverseTransformer(labelsCol)
+ .setInputCol(inputCol)
+ .setOutputCol(outputCol)
+ }
+}
+
+/**
+ * :: Experimental ::
+ * Transform a provided column back to the original input types using the
metadata on the
+ * labelsCol. Note: By default we keep the original columns during this
transformation,
+ * so the inverse should only be used on new columns such as predicted
labels.
+ */
+@Experimental
+class StringIndexerInverseTransformer private[ml] (
+ override val uid: String,
+ val labelsCol: String) extends Transformer
--- End diff --
Well, it should be...but it's true that many Models do not update the
output columns appropriately right now.
Maybe we should go back closer to what you had before, where
StringIndexerInverse takes a StringArrayParam called "labels" which lets the
user provide the lookup table? If it is not specified, then the input column
must have the metadata. If it is specified, then the input column metadata are
ignored. StringIndexerModel.invert() can set this parameter. Later on, it may
become unnecessary, but that's OK.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
