Implementation:Scikit learn Scikit learn RankingMetrics

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Machine Learning, Model Evaluation
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for computing ranking-based classification metrics provided by scikit-learn.

Description

The ranking metrics module provides functions to assess the performance of classification models based on predicted scores or rankings rather than hard class predictions. It includes widely used metrics such as Area Under the ROC Curve (AUC), average precision, precision-recall curves, ROC curves, and ranking-specific measures like NDCG and DCG. Functions named as *_score return values to maximize, while those named *_error or *_loss return values to minimize.

Usage

Use these metrics when evaluating classifiers that output probability estimates or decision scores, when you need threshold-independent evaluation, or when working with ranking and information retrieval tasks.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/metrics/_ranking.py

Signature

def auc(x, y)
def average_precision_score(y_true, y_score, *, average="macro", pos_label=1, sample_weight=None)
def det_curve(y_true, y_score, *, pos_label=None, sample_weight=None)
def roc_auc_score(y_true, y_score, *, average="macro", sample_weight=None, max_fpr=None, multi_class="raise", labels=None)
def precision_recall_curve(y_true, y_score, *, pos_label=None, sample_weight=None, drop_intermediate=False)
def roc_curve(y_true, y_score, *, pos_label=None, sample_weight=None, drop_intermediate=True)
def label_ranking_average_precision_score(y_true, y_score, *, sample_weight=None)
def coverage_error(y_true, y_score, *, sample_weight=None)
def label_ranking_loss(y_true, y_score, *, sample_weight=None)
def dcg_score(y_true, y_score, *, k=None, log_base=2, sample_weight=None, ignore_ties=False)
def ndcg_score(y_true, y_score, *, k=None, sample_weight=None, ignore_ties=False)
def top_k_accuracy_score(y_true, y_score, *, k=2, normalize=True, sample_weight=None, labels=None)
def confusion_matrix_at_thresholds(y_true, y_score, pos_label=None, sample_weight=None)

Import

from sklearn.metrics import auc, roc_auc_score, average_precision_score
from sklearn.metrics import precision_recall_curve, roc_curve, det_curve
from sklearn.metrics import ndcg_score, dcg_score, top_k_accuracy_score
from sklearn.metrics import label_ranking_average_precision_score, coverage_error, label_ranking_loss

I/O Contract

Inputs

Name	Type	Required	Description
y_true	array-like	Yes	Ground truth (correct) labels or binary label indicators
y_score	array-like	Yes	Target scores, probability estimates, or confidence values
sample_weight	array-like	No	Sample weights for weighted metric computation
average	str	No	Type of averaging for multiclass (macro, micro, weighted, samples)
pos_label	int, float, bool or str	No	The label of the positive class
k	int	No	Number of top elements to consider for DCG/NDCG/top-k metrics
max_fpr	float	No	Maximum false positive rate for partial AUC computation
multi_class	str	No	Multiclass handling strategy (raise, ovr, ovo)

Outputs

Name	Type	Description
score	float	Scalar metric value (for *_score functions)
fpr	ndarray	False positive rates (for curve functions)
tpr/recall	ndarray	True positive rates or recall values (for curve functions)
thresholds	ndarray	Decision thresholds used to compute curve points

Usage Examples

Basic Usage

import numpy as np
from sklearn.metrics import roc_auc_score, average_precision_score, roc_curve

# Binary classification scores
y_true = np.array([0, 0, 1, 1])
y_scores = np.array([0.1, 0.4, 0.35, 0.8])

# Compute ROC AUC
roc_auc = roc_auc_score(y_true, y_scores)
print(f"ROC AUC: {roc_auc:.3f}")

# Compute average precision
ap = average_precision_score(y_true, y_scores)
print(f"Average Precision: {ap:.3f}")

# Get ROC curve points
fpr, tpr, thresholds = roc_curve(y_true, y_scores)

# Compute AUC from curve
from sklearn.metrics import auc
roc_auc_from_curve = auc(fpr, tpr)

Related Pages

• Principle:Scikit_learn_Scikit_learn_Metric_Evaluation