Fix Binary Classification with different configurations. (#1463)

* Add default behaviour for Classification configuration.
Check for probas for RocAuc metrics and graphs.

* Fix linter

src/evidently/future/datasets.py

@@ -45,22 +45,81 @@ class ColumnInfo:
 
 @dataclasses.dataclass
 class BinaryClassification:
-    name: str = "default"
-    target: str = "target"
-    prediction_labels: Optional[str] = None
-    prediction_probas: Optional[str] = "prediction"
-    pos_label: Label = 1
-    labels: Optional[Dict[Label, str]] = None
+    name: str
+    target: str
+    prediction_labels: Optional[str]
+    prediction_probas: Optional[str]
+    pos_label: Label
+    labels: Optional[Dict[Label, str]]
+
+    def __init__(
+        self,
+        *,
+        name: str = "default",
+        target: Optional[str] = None,
+        prediction_labels: Optional[str] = None,
+        prediction_probas: Optional[str] = None,
+        pos_label: Optional[str] = None,
+        labels: Optional[Dict[Label, str]] = None,
+    ):
+        self.name = name
+        if (
+            target is None
+            and prediction_labels is None
+            and prediction_probas is None
+            and pos_label is None
+            and labels is None
+        ):
+            self.target = "target"
+            self.prediction_labels = None
+            self.prediction_probas = "prediction"
+            self.pos_label = 1
+            self.labels = None
+            return
+        if target is None or (prediction_labels is None and prediction_probas is None):
+            raise ValueError(
+                "Invalid BinaryClassification configuration:" " target and one of (labels or probas) should be set"
+            )
+        self.target = target
+        self.prediction_labels = prediction_labels
+        self.prediction_probas = prediction_probas
+        self.pos_label = pos_label if pos_label is not None else 1
+        self.labels = labels
 
 
 @dataclasses.dataclass
 class MulticlassClassification:
     name: str = "default"
     target: str = "target"
-    prediction_labels: str = "prediction"
+    prediction_labels: Optional[str] = "prediction"
     prediction_probas: Optional[List[str]] = None
     labels: Optional[Dict[Label, str]] = None
 
+    def __init__(
+        self,
+        *,
+        name: str = "default",
+        target: Optional[str] = None,
+        prediction_labels: Optional[str] = None,
+        prediction_probas: Optional[List[str]] = None,
+        labels: Optional[Dict[Label, str]] = None,
+    ):
+        self.name = name
+        if target is None and prediction_labels is None and prediction_probas is None and labels is None:
+            self.target = "target"
+            self.prediction_labels = "prediction"
+            self.prediction_probas = None
+            self.labels = None
+            return
+        if target is None or (prediction_labels is None and prediction_probas is None):
+            raise ValueError(
+                "Invalid MulticlassClassification configuration:" " target and one of (labels or probas) should be set"
+            )
+        self.target = target
+        self.prediction_labels = prediction_labels
+        self.prediction_probas = prediction_probas
+        self.labels = labels
+
 
 Classification = Union[BinaryClassification, MulticlassClassification]
 

src/evidently/future/presets/classification.py

@@ -103,12 +103,15 @@ def render(self, context: "Context", results: Dict[MetricId, MetricResult]) -> L
                 ClassificationConfusionMatrix(probas_threshold=self._probas_threshold),
                 _gen_classification_input_data,
             )[1]
-        if self._pr_curve:
+        classification = context.data_definition.get_classification("default")
+        if classification is None:
+            raise ValueError("Cannot use ClassificationQuality without a classification data")
+        if self._pr_curve and classification.prediction_probas is not None:
             render += context.get_legacy_metric(
                 ClassificationPRCurve(probas_threshold=self._probas_threshold),
                 _gen_classification_input_data,
             )[1]
-        if self._pr_table:
+        if self._pr_table and classification.prediction_probas is not None:
             render += context.get_legacy_metric(
                 ClassificationPRTable(probas_threshold=self._probas_threshold),
                 _gen_classification_input_data,
@@ -122,12 +125,20 @@ def __init__(self, probas_threshold: Optional[float] = None, k: Optional[int] =
         self._k = k
 
     def generate_metrics(self, context: "Context") -> List[Metric]:
+        classification = context.data_definition.get_classification("default")
+        if classification is None:
+            raise ValueError("Cannot use ClassificationPreset without a classification configration")
         return [
             F1ByLabel(probas_threshold=self._probas_threshold, k=self._k),
             PrecisionByLabel(probas_threshold=self._probas_threshold, k=self._k),
             RecallByLabel(probas_threshold=self._probas_threshold, k=self._k),
-            RocAucByLabel(probas_threshold=self._probas_threshold, k=self._k),
-        ]
+        ] + (
+            []
+            if classification.prediction_probas is None
+            else [
+                RocAucByLabel(probas_threshold=self._probas_threshold, k=self._k),
+            ]
+        )
 
     def render(self, context: "Context", results: Dict[MetricId, MetricResult]):
         render = context.get_legacy_metric(