RWkNN

src/capymoa/classifier/__init__.py

@@ -18,6 +18,7 @@
 from ._samknn import SAMkNN
 from ._dynamic_weighted_majority import DynamicWeightedMajority
 from ._csmote import CSMOTE
+from ._rwknn import RWkNN
 
 __all__ = [
     "AdaptiveRandomForestClassifier",
@@ -39,5 +40,6 @@
     "HoeffdingAdaptiveTree",
     "SAMkNN",
     "DynamicWeightedMajority",
-    "CSMOTE"
+    "CSMOTE",
+    "RWkNN"
 ]

src/capymoa/classifier/_rwknn.py

@@ -0,0 +1,72 @@
+from capymoa.base import MOAClassifier
+from moa.classifiers.lazy import RW_kNN as _MOA_RWkNN
+from capymoa.stream import Schema
+from capymoa._utils import build_cli_str_from_mapping_and_locals
+
+
+class RWkNN(MOAClassifier):
+    """RwkNN
+
+    Reference:
+
+    'Incremental k-Nearest Neighbors Using Reservoir Sampling for Data Streams
+    Maroua Bahri, Albert Bifet
+    Discovery Science: 24th International Conference, 2021
+    <https://link.springer.com/chapter/10.1007/978-3-030-88942-5_10>`_
+
+    Example usages:
+
+    >>> from capymoa.datasets import ElectricityTiny
+    >>> from capymoa.classifier import RWkNN
+    >>> from capymoa.evaluation import prequential_evaluation
+    >>> stream = ElectricityTiny()
+    >>> schema = stream.get_schema()
+    >>> learner = RWkNN(schema)
+    >>> results = prequential_evaluation(stream, learner, max_instances=1000)
+    >>> results["cumulative"].accuracy()
+    81.8
+    """
+
+
+    def __init__(
+        self,
+        schema: Schema,
+        k: int = 5,
+        limitW: int = 500,
+        limitR: int = 500,
+        nearest_neighbor_search: str = "LinearNN" 
+    ):
+
+        """ RW KNN Classifier
+
+        :param schema: The schema of the stream.
+        :param k: The number of neighbors.
+        :param limit_w: The maximum number of instances to store in the window.
+        :param limit_r: The maximum number of instances to store in the reservoir.
+        :param nearest_neighbor_search: Nearest Neighbour Search to use.
+        """
+
+        self.nearest_neighbor_search = nearest_neighbor_search
+        if isinstance(self.nearest_neighbor_search, str) and (nearest_neighbor_search == "LinearNN" or nearest_neighbor_search == "KDTree"):
+            self.nearest_neighbor_search = nearest_neighbor_search
+        else:
+            # Raise an exception with information about valid options for max_features
+            raise ValueError("Invalid value for nearest_neighbor_search. Valid options: LinearNN, KDTree")
+
+
+        mapping = {
+            "k": "-k",
+            "limitW": "-w",
+            "limitR": "-r",
+            "nearest_neighbor_search": "-n",
+        }
+
+
+        config_str = build_cli_str_from_mapping_and_locals(mapping, locals())
+        self.moa_learner = _MOA_RWkNN()
+        super(RWkNN, self).__init__(
+            schema=schema,
+            CLI=config_str,
+            moa_learner=self.moa_learner,
+        )
+

tests/test_classifiers.py

@@ -18,7 +18,9 @@
     CSMOTE,
     LeveragingBagging,
     OnlineAdwinBagging,
+    RWkNN
 )
+
 from capymoa.base import Classifier
 from capymoa.base import MOAClassifier
 from capymoa.datasets import ElectricityTiny
@@ -62,6 +64,7 @@
         (partial(CSMOTE), 80.55, 79.0, None),
         (partial(LeveragingBagging), 86.7, 91.0, None),
         (partial(OnlineAdwinBagging), 85.25, 92.0, None),
+        (partial(RWkNN), 85.25, 92.0, None),
     ],
     ids=[
         "OnlineBagging",
@@ -77,6 +80,7 @@
         "OzaBoost",
         "MajorityClass",
         "NoChange",
+        "OnlineAccuracyUpdatedEnsemble",
         "OnlineSmoothBoost",
         "StreamingRandomPatches",
         "HoeffdingAdaptiveTree",
@@ -85,6 +89,7 @@
         "CSMOTE",
         "LeveragingBagging",
         "OnlineAdwinBagging",
+        "RWkNN"
     ],
 )
 def test_classifiers(