Add data drift method Mann-Whitney U test

Author: jaime-cespedes-sisniega

README.md

@@ -284,8 +284,8 @@ The currently implemented detectors are listed in the following table.
     <td style="text-align: center; border: 1px solid grey; padding: 8px;"><a href="https://doi.org/10.1007/978-3-540-75488-6_27">Nishida and Yamauchi (2007)</a></td>
   </tr>
   <tr>
-    <td rowspan="14" style="text-align: center; border: 1px solid grey; padding: 8px;">Data drift</td>
-    <td rowspan="12" style="text-align: center; border: 1px solid grey; padding: 8px;">Batch</td>
+    <td rowspan="15" style="text-align: center; border: 1px solid grey; padding: 8px;">Data drift</td>
+    <td rowspan="13" style="text-align: center; border: 1px solid grey; padding: 8px;">Batch</td>
     <td rowspan="8" style="text-align: center; border: 1px solid grey; padding: 8px;">Distance based</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">U</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">N</td>
@@ -335,7 +335,7 @@ The currently implemented detectors are listed in the following table.
     <td style="text-align: center; border: 1px solid grey; padding: 8px;"><a href="https://doi.org/10.1057/jors.2008.144">Wu and Olson (2010)</a></td>
   </tr>
   <tr>
-    <td rowspan="4" style="text-align: center; border: 1px solid grey; padding: 8px;">Statistical test</td>
+    <td rowspan="5" style="text-align: center; border: 1px solid grey; padding: 8px;">Statistical test</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">U</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">C</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">Chi-square test</td>
@@ -353,6 +353,12 @@ The currently implemented detectors are listed in the following table.
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">Kolmogorov-Smirnov test</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;"><a href="https://doi.org/10.2307/2280095">Massey Jr (1951)</a></td>
   </tr>
+  <tr>
+    <td style="text-align: center; border: 1px solid grey; padding: 8px;">U</td>
+    <td style="text-align: center; border: 1px solid grey; padding: 8px;">N</td>
+    <td style="text-align: center; border: 1px solid grey; padding: 8px;">Mann-Whitney U test</td>
+    <td style="text-align: center; border: 1px solid grey; padding: 8px;"><a href="https://doi.org/10.1214/aoms/1177730491">Mann and Whitney (1947)</a></td>
+  </tr>
   <tr>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">U</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">N</td>

docs/source/api_reference/detectors/data_drift/batch.md

@@ -50,5 +50,6 @@ The {mod}`frouros.detectors.data_drift.batch` module contains batch data drift d
     ChiSquareTest
     CVMTest
     KSTest
+    MannWhitneyUTest
     WelchTTest
 ```

frouros/detectors/data_drift/__init__.py

@@ -11,6 +11,7 @@
     KL,
     KSTest,
     PSI,
+    MannWhitneyUTest,
     MMD,
     WelchTTest,
 )
@@ -29,6 +30,7 @@
     "KL",
     "KSTest",
     "PSI",
+    "MannWhitneyUTest",
     "MMDStreaming",
     "WelchTTest",
 ]

frouros/detectors/data_drift/batch/__init__.py

@@ -14,6 +14,7 @@
     ChiSquareTest,
     CVMTest,
     KSTest,
+    MannWhitneyUTest,
     WelchTTest,
 )
 
@@ -28,6 +29,7 @@
     "KL",
     "KSTest",
     "PSI",
+    "MannWhitneyUTest",
     "MMD",
     "WelchTTest",
 ]

frouros/detectors/data_drift/batch/statistical_test/__init__.py

@@ -3,11 +3,13 @@
 from .chisquare import ChiSquareTest
 from .cvm import CVMTest
 from .ks import KSTest
+from .mann_whitney_u import MannWhitneyUTest
 from .welch_t_test import WelchTTest
 
 __all__ = [
     "ChiSquareTest",
     "CVMTest",
     "KSTest",
+    "MannWhitneyUTest",
     "WelchTTest",
 ]

frouros/detectors/data_drift/batch/statistical_test/mann_whitney_u.py

@@ -0,0 +1,49 @@
+"""Mann-Whitney U test module."""
+
+from typing import Optional, List, Union
+
+import numpy as np  # type: ignore
+from scipy.stats import mannwhitneyu  # type: ignore
+
+from frouros.callbacks.batch.base import BaseCallbackBatch
+from frouros.detectors.data_drift.base import NumericalData, UnivariateData
+from frouros.detectors.data_drift.batch.statistical_test.base import (
+    BaseStatisticalTest,
+    StatisticalResult,
+)
+
+
+class MannWhitneyUTest(BaseStatisticalTest):
+    """Mann–Whitney U test [mann1947test]_ detector.
+
+    :References:
+
+    .. [mann1947test] Mann, Henry B., and Donald R. Whitney.
+        "On a test of whether one of two random variables is stochastically larger than
+        the other."
+        The annals of mathematical statistics (1947): 50-60.
+    """
+
+    def __init__(
+        self,
+        callbacks: Optional[Union[BaseCallbackBatch, List[BaseCallbackBatch]]] = None,
+    ) -> None:
+        """Init method.
+
+        :param callbacks: callbacks
+        :type callbacks: Optional[Union[BaseCallbackBatch, List[BaseCallbackBatch]]]
+        """
+        super().__init__(
+            data_type=NumericalData(),
+            statistical_type=UnivariateData(),
+            callbacks=callbacks,
+        )
+
+    def _statistical_test(
+        self, X_ref: np.ndarray, X: np.ndarray, **kwargs  # noqa: N803
+    ) -> StatisticalResult:
+        test = mannwhitneyu(
+            x=X_ref, y=X, alternative="two-sided", nan_policy="raise", **kwargs
+        )
+        test = StatisticalResult(statistic=test.statistic, p_value=test.pvalue)
+        return test

frouros/tests/integration/test_data_drift.py

@@ -19,6 +19,7 @@
     ChiSquareTest,
     CVMTest,
     KSTest,
+    MannWhitneyUTest,
     WelchTTest,
 )
 from frouros.detectors.data_drift.batch.base import BaseDataDriftBatch
@@ -161,6 +162,7 @@ def test_batch_distance_bins_based_univariate_same_distribution(
     [
         (CVMTest(), 3776.09848103, 5.38105056e-07),
         (KSTest(), 0.99576271, 0.0),
+        (MannWhitneyUTest(), 6912.0, 0.0),
         (WelchTTest(), -287.92032554, 0.0),
     ],
 )