From 4e080928a8b6d62e57f5a08e7637790f26b76ace Mon Sep 17 00:00:00 2001
From: John Wittenauer <jdwittenauer@gmail.com>
Date: Fri, 15 May 2015 21:19:39 -0400
Subject: [PATCH] Added classes_ attribute to scikit-learn wrapper.

---
 wrapper/xgboost.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/wrapper/xgboost.py b/wrapper/xgboost.py
index 7e3fa2dc4..46a229cd6 100644
--- a/wrapper/xgboost.py
+++ b/wrapper/xgboost.py
@@ -935,8 +935,8 @@ class XGBClassifier(XGBModel, XGBClassifier):
                                             base_score, seed)
 
     def fit(self, X, y, sample_weight=None):
-        y_values = list(np.unique(y))
-        self.n_classes_ = len(y_values)
+        self.classes_ = list(np.unique(y))
+        self.n_classes_ = len(self.classes_)
         if self.n_classes_ > 2:
             # Switch to using a multiclass objective in the underlying XGB instance
             self.objective = "multi:softprob"