python
16 lines · 1 tab
Dr. Elena Vasquez
Apr 2026
1 tab
from sklearn.metrics import (
average_precision_score,
classification_report,
precision_recall_curve,
roc_auc_score,
)
probabilities = model.predict_proba(X_valid)[:, 1]
predictions = (probabilities >= 0.35).astype(int)
print('ROC AUC:', roc_auc_score(y_valid, probabilities))
print('PR AUC:', average_precision_score(y_valid, probabilities))
print(classification_report(y_valid, predictions, digits=3))
precision, recall, thresholds = precision_recall_curve(y_valid, probabilities)
print('threshold samples:', list(zip(thresholds[:5], precision[:5], recall[:5])))
1 file · python
Explain with highlit
Accuracy is a bad comfort metric when the positive class is rare. I care more about precision, recall, PR AUC, calibration, and how thresholding changes operational workload. The right metric depends on the cost of false negatives versus false positives, not on what is easiest to explain on a slide.
Related snips
python
import numpy as np
from sklearn.metrics import confusion_matrix
probabilities = model.predict_proba(X_valid)[:, 1]
thresholds = np.linspace(0.1, 0.9, 9)Confusion matrix diagnostics for threshold selection
confusion-matrix
thresholding
evaluation
by Dr. Elena Vasquez
1 tab
python
from sklearn.model_selection import StratifiedKFold, train_test_split, cross_validate
from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegressionTrain test split and stratified cross validation done properly
cross-validation
evaluation
scikit-learn
by Dr. Elena Vasquez
1 tab
Share this code
Here's the card — post it anywhere.
