Use raw auc then normalize for detection test

.gitignore

@@ -109,3 +109,5 @@ ENV/
 # OS Files
 .DS_Store
 
+# vcode stuff
+.vcode/

sdmetrics/single_table/__init__.py

@@ -6,7 +6,8 @@
 from sdmetrics.single_table.bayesian_network import BNLikelihood, BNLogLikelihood
 from sdmetrics.single_table.detection.base import DetectionMetric
 from sdmetrics.single_table.detection.sklearn import (
-    LogisticDetection, ScikitLearnClassifierDetectionMetric, SVCDetection)
+    GradientBoostingDetection, LogisticDetection, ScikitLearnClassifierDetectionMetric,
+    SVCDetection)
 from sdmetrics.single_table.efficacy.base import MLEfficacyMetric
 from sdmetrics.single_table.efficacy.binary import (
     BinaryAdaBoostClassifier, BinaryDecisionTreeClassifier, BinaryEfficacyMetric,
@@ -47,6 +48,7 @@
     'DetectionMetric',
     'LogisticDetection',
     'SVCDetection',
+    'GradientBoostingDetection',
     'ScikitLearnClassifierDetectionMetric',
     'MLEfficacyMetric',
     'BinaryEfficacyMetric',

sdmetrics/single_table/detection/__init__.py

@@ -1,8 +1,10 @@
 """Machine Learning Detection metrics for single table datasets."""
 
-from sdmetrics.single_table.detection.sklearn import LogisticDetection, SVCDetection
+from sdmetrics.single_table.detection.sklearn import (
+    GradientBoostingDetection, LogisticDetection, SVCDetection)
 
 __all__ = [
+    'GradientBoostingDetection',
     'LogisticDetection',
     'SVCDetection'
 ]

sdmetrics/single_table/detection/base.py

@@ -34,7 +34,7 @@ class DetectionMetric(SingleTableMetric):
     """
 
     name = 'SingleTable Detection'
-    goal = Goal.MAXIMIZE
+    goal = Goal.MINIMIZE
     min_value = 0.0
     max_value = 1.0
 
@@ -50,7 +50,7 @@ def compute(cls, real_data, synthetic_data, metadata=None):
         This builds a Machine Learning Classifier that learns to tell the synthetic
         data apart from the real data, which later on is evaluated using Cross Validation.
 
-        The output of the metric is one minus the average ROC AUC score obtained.
+        The output of the metric is the average ROC AUC score obtained.
 
         Args:
             real_data (Union[numpy.ndarray, pandas.DataFrame]):
@@ -85,22 +85,24 @@ def compute(cls, real_data, synthetic_data, metadata=None):
                 y_pred = cls._fit_predict(X[train_index], y[train_index], X[test_index])
                 roc_auc = roc_auc_score(y[test_index], y_pred)
 
-                scores.append(max(0.5, roc_auc) * 2 - 1)
+                scores.append(max(0.5,roc_auc))
 
-            return 1 - np.mean(scores)
+            return np.mean(scores)
         except ValueError as err:
             raise IncomputableMetricError(f'DetectionMetric: Unable to be fit with error {err}')
 
     @classmethod
     def normalize(cls, raw_score):
-        """Return the `raw_score` as is, since it is already normalized.
+        """Return the `raw_score`normalized to be higher-is-better in [0,1]
 
         Args:
             raw_score (float):
                 The value of the metric from `compute`.
 
         Returns:
             float:
-                Simply returns `raw_score`.
+                Returns `2*(1-raw_score)`.
         """
-        return super().normalize(raw_score)
+        assert raw_score >= 0.5, "raw auc score should be in [0.5,1]"
+        score = 2 * (1 - raw_score)
+        return super().normalize(score)

sdmetrics/single_table/detection/sklearn.py

@@ -1,5 +1,6 @@
 """scikit-learn based DetectionMetrics for single table datasets."""
 
+from sklearn.ensemble import GradientBoostingClassifier
 from sklearn.impute import SimpleImputer
 from sklearn.linear_model import LogisticRegression
 from sklearn.pipeline import Pipeline
@@ -67,3 +68,19 @@ class SVCDetection(ScikitLearnClassifierDetectionMetric):
     @staticmethod
     def _get_classifier():
         return SVC(probability=True, gamma='scale')
+
+
+class GradientBoostingDetection(ScikitLearnClassifierDetectionMetric):
+    """ScikitLearnClassifierDetectionMetric based on a GradientBoostingClassifier.
+
+    This metric builds a GradientBoostingClassifier Classifier that learns to tell the synthetic
+    data apart from the real data, which later on is evaluated using Cross Validation.
+
+    The output of the metric is one minus the average ROC AUC score obtained.
+    """
+
+    name = 'GradientBoosting Detection'
+
+    @staticmethod
+    def _get_classifier():
+        return GradientBoostingClassifier()

tests/integration/single_table/test_single_table.py

@@ -7,7 +7,8 @@
 from sdmetrics.goal import Goal
 from sdmetrics.single_table.base import SingleTableMetric
 from sdmetrics.single_table.bayesian_network import BNLikelihood, BNLogLikelihood
-from sdmetrics.single_table.detection import LogisticDetection, SVCDetection
+from sdmetrics.single_table.detection import (
+    GradientBoostingDetection, LogisticDetection, SVCDetection)
 from sdmetrics.single_table.multi_column_pairs import (
     ContingencySimilarity, ContinuousKLDivergence, DiscreteKLDivergence)
 from sdmetrics.single_table.multi_single_column import (
@@ -17,6 +18,7 @@
 METRICS = [
     CSTest,
     KSComplement,
+    GradientBoostingDetection,
     LogisticDetection,
     SVCDetection,
     ContinuousKLDivergence,