Implementation:Evidentlyai Evidently Task Configuration Classes

Knowledge Sources	Evidently Evidently Data Definition Guide
Domains	ML_Evaluation, Classification, Regression
Last Updated	2026-02-14 12:00 GMT

Overview

Concrete dataclass configurations for mapping ML task columns provided by the Evidently library.

Description

Three dataclass-based configuration classes that map ML task columns to Evidently's evaluation engine:

• BinaryClassification: Maps target, predicted labels, predicted probabilities, and positive label for binary classification tasks
• MulticlassClassification: Maps target, predicted labels, and per-class probability columns for multiclass tasks
• Regression: Maps target and prediction columns for regression tasks

These classes are passed as lists to DataDefinition(classification=[...], regression=[...]) to enable task-specific metric computation.

Usage

Import these classes when you need to evaluate ML model quality using Evidently's classification or regression presets. Use them as parameters to DataDefinition.

Code Reference

Source Location

• Repository: evidently
• File: src/evidently/core/datasets.py
• Lines: L85-239

Signature

@dataclasses.dataclass
class BinaryClassification:
    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,
    ):
        """
        Args:
            name: Identifier for this classification task.
            target: Column name with true binary labels.
            prediction_labels: Column name with predicted binary labels.
            prediction_probas: Column name with predicted probabilities.
            pos_label: Value representing the positive class.
            labels: Optional mapping of label values to display names.
        """

@dataclasses.dataclass
class MulticlassClassification:
    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,
    ):
        """
        Args:
            name: Identifier for this classification task.
            target: Column name with true class labels.
            prediction_labels: Column name with predicted class labels.
            prediction_probas: List of column names with per-class probabilities.
            labels: Optional mapping of label values to display names.
        """

@dataclasses.dataclass
class Regression:
    name: str = "default"
    target: str = "target"
    prediction: str = "prediction"

Import

from evidently.core.datasets import BinaryClassification, MulticlassClassification, Regression

I/O Contract

Inputs (BinaryClassification)

Name	Type	Required	Description
name	str	No	Task identifier (default: "default")
target	Optional[str]	Conditional	Column with true labels (required if any param set)
prediction_labels	Optional[str]	Conditional	Column with predicted labels
prediction_probas	Optional[str]	Conditional	Column with predicted probabilities
pos_label	Optional[str]	No	Value for positive class (default: 1)
labels	Optional[Dict]	No	Label value to display name mapping

Inputs (Regression)

Name	Type	Required	Description
name	str	No	Task identifier (default: "default")
target	str	No	Column with actual values (default: "target")
prediction	str	No	Column with predicted values (default: "prediction")

Outputs

Name	Type	Description
BinaryClassification	BinaryClassification	Config object passed to DataDefinition(classification=[...])
MulticlassClassification	MulticlassClassification	Config object passed to DataDefinition(classification=[...])
Regression	Regression	Config object passed to DataDefinition(regression=[...])

Usage Examples

Binary Classification Setup

from evidently import DataDefinition
from evidently.core.datasets import BinaryClassification

# Map columns for fraud detection binary classification
data_def = DataDefinition(
    classification=[
        BinaryClassification(
            target="is_fraud",
            prediction_labels="predicted_fraud",
            prediction_probas="fraud_probability",
            pos_label=1,
        )
    ]
)

Multiclass Classification Setup

from evidently.core.datasets import MulticlassClassification

data_def = DataDefinition(
    classification=[
        MulticlassClassification(
            target="species",
            prediction_labels="predicted_species",
            prediction_probas=["prob_setosa", "prob_versicolor", "prob_virginica"],
        )
    ]
)

Regression Setup

from evidently.core.datasets import Regression

data_def = DataDefinition(
    regression=[
        Regression(target="price", prediction="predicted_price")
    ]
)

Related Pages

Implements Principle

• Principle:Evidentlyai_Evidently_ML_Task_Configuration