Implementation:Evidentlyai Evidently ClassificationQuality Preset

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

Overview

Concrete preset container for computing classification quality metrics provided by the Evidently library.

Description

ClassificationQuality is a MetricContainer that expands into individual classification metrics (Accuracy, Precision, Recall, F1Score, RocAuc, LogLoss, TPR, TNR, FPR, FNR) plus optional visualizations (confusion matrix, PR curve, PR table). Each metric can have optional test conditions.

Usage

Include in a Report metrics list when evaluating classification model performance. Requires DataDefinition with classification task.

Code Reference

Source Location

• Repository: evidently
• File: src/evidently/presets/classification.py
• Lines: L46-120

Signature

class ClassificationQuality(MetricContainer):
    def __init__(
        self,
        classification_name: str = "default",
        probas_threshold: Optional[float] = None,
        conf_matrix: bool = False,
        pr_curve: bool = False,
        pr_table: bool = False,
        accuracy_tests: GenericSingleValueMetricTests = None,
        precision_tests: GenericSingleValueMetricTests = None,
        recall_tests: GenericSingleValueMetricTests = None,
        f1score_tests: GenericSingleValueMetricTests = None,
        rocauc_tests: GenericSingleValueMetricTests = None,
        logloss_tests: GenericSingleValueMetricTests = None,
        tpr_tests: GenericSingleValueMetricTests = None,
        tnr_tests: GenericSingleValueMetricTests = None,
        fpr_tests: GenericSingleValueMetricTests = None,
        fnr_tests: GenericSingleValueMetricTests = None,
        include_tests: bool = True,
    ):

Import

from evidently.presets import ClassificationQuality

I/O Contract

Inputs

Name	Type	Required	Description
classification_name	str	No	Task name matching DataDefinition (default: "default")
probas_threshold	Optional[float]	No	Probability threshold for binary classification
conf_matrix	bool	No	Show confusion matrix (default: False)
pr_curve	bool	No	Show PR curve (default: False)
include_tests	bool	No	Enable auto-tests (default: True)

Outputs (Expanded Metrics)

Metric	Type	Description
Accuracy	float	Overall classification accuracy
Precision	float	Positive predictive value
Recall	float	True positive rate / sensitivity
F1Score	float	Harmonic mean of precision and recall
RocAuc	float	Area under ROC curve
LogLoss	float	Log loss / cross-entropy loss

Usage Examples

Basic Classification Report

from evidently import Report, Dataset, DataDefinition
from evidently.core.datasets import BinaryClassification
from evidently.presets import ClassificationQuality

data_def = DataDefinition(
    classification=[BinaryClassification(target="target", prediction_labels="prediction")]
)

report = Report([ClassificationQuality()], include_tests=True)
snapshot = report.run(
    Dataset.from_pandas(df_current, data_def),
    Dataset.from_pandas(df_reference, data_def),
)
snapshot.save_html("classification_report.html")

Related Pages

Implements Principle

• Principle:Evidentlyai_Evidently_Classification_Quality_Evaluation