FIX use balanced accuracy from scikit-learn

rampwf/score_types/balanced_accuracy.py

@@ -1,27 +1,44 @@
-"""Balanced accuracy.
+import warnings
 
-From https://github.com/ch-imad/AutoMl_Challenge/blob/2353ec0/Starting_kit/scoring_program/libscores.py#L187  # noqa
+import numpy as np
+
+from sklearn.metrics import confusion_matrix
+from sklearn.metrics.classification import _check_targets
 
-See the thread at 
-https://github.com/rhiever/tpot/issues/108#issuecomment-317067760
-about the different definitions.
-"""
 from .classifier_base import ClassifierBaseScoreType
-from .macro_averaged_recall import MacroAveragedRecall
+
+
+def _balanced_accuracy_score(y_true, y_pred, sample_weight=None,
+                             adjusted=True):
+    """FIXME: port implementation of balanced accuracy from scikit-learn 0.20.
+    """
+    C = confusion_matrix(y_true, y_pred, sample_weight=sample_weight)
+    with np.errstate(divide='ignore', invalid='ignore'):
+        per_class = np.diag(C) / C.sum(axis=1)
+    if np.any(np.isnan(per_class)):
+        warnings.warn('y_pred contains classes not in y_true')
+        per_class = per_class[~np.isnan(per_class)]
+    score = np.mean(per_class)
+    if adjusted:
+        n_classes = len(per_class)
+        chance = 1 / n_classes
+        score -= chance
+        score /= 1 - chance
+    return score
 
 
 class BalancedAccuracy(ClassifierBaseScoreType):
     is_lower_the_better = False
     minimum = 0.0
     maximum = 1.0
 
-    def __init__(self, name='balanced_accuracy', precision=2):
+    def __init__(self, name='balanced_accuracy', precision=2, adjusted=True):
         self.name = name
         self.precision = precision
+        self.adjusted = adjusted
 
     def __call__(self, y_true_label_index, y_pred_label_index):
-        mac = MacroAveragedRecall()
-        tpr = mac(y_true_label_index, y_pred_label_index)
-        base_tpr = 1. / len(self.label_names)
-        score = (tpr - base_tpr) / (1 - base_tpr)
+        score = _balanced_accuracy_score(y_true_label_index,
+                                         y_pred_label_index,
+                                         adjusted=self.adjusted)
         return score