feat: factory pattern for creating datasets

aequitas/datasets/binary_label_dataset.py

@@ -0,0 +1,20 @@
+from aequitas.datasets.structured_dataset import StructuredDataset
+from aequitas.metrics.binary_label_dataset_metric import BinaryLabelDatasetScoresMetric
+
+
+class BinaryLabelDataset(StructuredDataset):
+
+    def __init__(self, favorable_label=1., unfavorable_label=0., **kwargs):
+        self._favorable_label = float(favorable_label)
+        self._unfavorable_label = float(unfavorable_label)
+
+        super(BinaryLabelDataset, self).__init__(**kwargs)
+
+    @classmethod
+    def new_instance(cls):
+        return cls
+
+    @property
+    def metrics(self, **kwargs):
+        return BinaryLabelDatasetScoresMetric(**kwargs)
+

aequitas/datasets/concrete_dataset_factories.py

@@ -0,0 +1,12 @@
+from aequitas.datasets.dataset_factory import DatasetFactory
+from aequitas.datasets.binary_label_dataset import BinaryLabelDataset
+
+
+class BinaryLabelDatasetFactory(DatasetFactory):
+    def __init__(self):
+        pass
+
+    def create_dataset(self, **kwargs) -> BinaryLabelDataset:
+        print("DEBUG:calling BinaryLabelDataset constructor")
+        ds = BinaryLabelDataset(favorable_label=1., unfavorable_label=0., **kwargs)
+        return ds

aequitas/datasets/create_dataset.py

@@ -0,0 +1,12 @@
+from aequitas.datasets.structured_dataset import StructuredDataset
+from aequitas.datasets.concrete_dataset_factories import BinaryLabelDatasetFactory
+
+
+class CreateDataset:
+    def __init__(self, dataset_type):
+        self.dataset_type = dataset_type
+
+    def create_dataset(self, **kwargs) -> StructuredDataset:
+        if self.dataset_type == "binary label":
+            bldf = BinaryLabelDatasetFactory()
+            return bldf.create_dataset(**kwargs)

aequitas/datasets/dataset_factory.py

@@ -0,0 +1,10 @@
+from abc import ABC, abstractmethod
+from aequitas.datasets.structured_dataset import StructuredDataset
+
+
+class DatasetFactory(ABC):
+    @abstractmethod
+    def create_dataset(self, **kwargs) -> StructuredDataset:
+        raise NotImplementedError
+
+

aequitas/datasets/structured_dataset.py

@@ -15,15 +15,7 @@ def __init__(self,
         kwargs["df"] = self._df
         super(StructuredDataset, self).__init__(**kwargs)
 
-    @property
-    def strategy(self):
-        return self.__strategy
-
     @property
     @abstractmethod
     def metrics(self):
         raise NotImplementedError
-
-    def __custom_preprocessing(self, df: pd.DataFrame) -> pd.DataFrame:
-        print("Applying custom preprocessing")
-        return self.strategy.custom_preprocessing(df)