[Doc] Document that AUC and AUCPR are for binary classification/ranki…

…ng [skip ci] (#5899)

doc/parameter.rst

@@ -399,8 +399,8 @@ Specify the learning task and the corresponding learning objective. The objectiv
     - ``error@t``: a different than 0.5 binary classification threshold value could be specified by providing a numerical value through 't'.
     - ``merror``: Multiclass classification error rate. It is calculated as ``#(wrong cases)/#(all cases)``.
     - ``mlogloss``: `Multiclass logloss <http://scikit-learn.org/stable/modules/generated/sklearn.metrics.log_loss.html>`_.
-    - ``auc``: `Area under the curve <http://en.wikipedia.org/wiki/Receiver_operating_characteristic#Area_under_curve>`_
-    - ``aucpr``: `Area under the PR curve <https://en.wikipedia.org/wiki/Precision_and_recall>`_
+    - ``auc``: `Area under the curve <http://en.wikipedia.org/wiki/Receiver_operating_characteristic#Area_under_curve>`_. Available for binary classification and learning-to-rank tasks.
+    - ``aucpr``: `Area under the PR curve <https://en.wikipedia.org/wiki/Precision_and_recall>`_. Available for binary classification and learning-to-rank tasks.
     - ``ndcg``: `Normalized Discounted Cumulative Gain <http://en.wikipedia.org/wiki/NDCG>`_
     - ``map``: `Mean Average Precision <http://en.wikipedia.org/wiki/Mean_average_precision#Mean_average_precision>`_
     - ``ndcg@n``, ``map@n``: 'n' can be assigned as an integer to cut off the top positions in the lists for evaluation.