Spaces:
Build error
Build error
File size: 2,919 Bytes
583e46a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 | """
Evaluation metrics module.
Computes accuracy, precision, recall, F1 (macro and per-class).
"""
from typing import Optional
import numpy as np
from sklearn.metrics import (
accuracy_score,
precision_score,
recall_score,
f1_score,
classification_report,
confusion_matrix,
)
def compute_metrics(
y_true: list[int],
y_pred: list[int],
class_names: Optional[list[str]] = None,
average: str = 'macro',
) -> dict:
"""
Compute comprehensive classification metrics.
Returns:
dict with accuracy, macro/weighted F1, precision, recall,
per-class metrics, and classification report.
"""
y_true = np.array(y_true)
y_pred = np.array(y_pred)
metrics = {
'accuracy': float(accuracy_score(y_true, y_pred)),
'macro_precision': float(precision_score(y_true, y_pred, average='macro', zero_division=0)),
'macro_recall': float(recall_score(y_true, y_pred, average='macro', zero_division=0)),
'macro_f1': float(f1_score(y_true, y_pred, average='macro', zero_division=0)),
'weighted_precision': float(precision_score(y_true, y_pred, average='weighted', zero_division=0)),
'weighted_recall': float(recall_score(y_true, y_pred, average='weighted', zero_division=0)),
'weighted_f1': float(f1_score(y_true, y_pred, average='weighted', zero_division=0)),
}
# Per-class metrics
per_class_precision = precision_score(y_true, y_pred, average=None, zero_division=0)
per_class_recall = recall_score(y_true, y_pred, average=None, zero_division=0)
per_class_f1 = f1_score(y_true, y_pred, average=None, zero_division=0)
per_class = {}
for i, f1_val in enumerate(per_class_f1):
name = class_names[i] if class_names and i < len(class_names) else str(i)
per_class[name] = {
'precision': float(per_class_precision[i]),
'recall': float(per_class_recall[i]),
'f1': float(f1_val),
'support': int(np.sum(y_true == i)),
}
metrics['per_class'] = per_class
# Confusion matrix
cm = confusion_matrix(y_true, y_pred)
metrics['confusion_matrix'] = cm.tolist()
# Classification report string
if class_names:
report = classification_report(
y_true, y_pred, target_names=class_names, zero_division=0,
)
else:
report = classification_report(y_true, y_pred, zero_division=0)
metrics['classification_report'] = report
return metrics
def compute_top_k_accuracy(
probabilities: list[list[float]],
y_true: list[int],
k: int = 3,
) -> float:
"""Compute top-k accuracy from probability distributions."""
correct = 0
for probs, true_label in zip(probabilities, y_true):
top_k_preds = np.argsort(probs)[-k:]
if true_label in top_k_preds:
correct += 1
return correct / len(y_true) if y_true else 0.0
|