Skip to content

📌 Purpose

This companion guide provides a visual and interpretive framework for evaluating logistic regression models. It centralizes best-practice plots, metrics, and interpretation notes to support clear, consistent model evaluation workflows.

🧩 Confusion Matrix

📈 Visual:

  • 2×2 table showing True Positive (TP), False Positive (FP), False Negative (FN), True Negative (TN)
  • Heatmap style preferred
from sklearn.metrics import confusion_matrix
import seaborn as sns
cm = confusion_matrix(y_test, y_pred)
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues')

🧠 Interpretation:

Cell Meaning
TP Correctly predicted positive
FP Incorrectly predicted positive
FN Missed positive
TN Correctly predicted negative
  • Diagonal dominance (high TP/TN) = good model
  • High FP = bad for precision
  • High FN = bad for recall

📋 Classification Report

📈 Visual:

  • Text output or table summarizing:
  • Precision
  • Recall
  • F1-score
  • Support (number of examples per class)
from sklearn.metrics import classification_report
print(classification_report(y_test, y_pred))

🧠 Interpretation:

Metric Insight
Precision Trustworthiness of positive predictions
Recall Coverage of actual positives
F1-score Balanced metric if classes are imbalanced
  • Low precision: many false alarms
  • Low recall: many missed positives

📈 ROC Curve + AUC

📈 Visual:

  • Plot of True Positive Rate vs False Positive Rate
  • AUC = area under the ROC curve
from sklearn.metrics import roc_curve, roc_auc_score
fpr, tpr, thresholds = roc_curve(y_test, clf.predict_proba(X_test)[:,1])
plt.plot(fpr, tpr)
plt.plot([0, 1], [0, 1], 'k--')  # Random guess line

🧠 Interpretation:

AUC Value Meaning
~0.5 Random guessing
0.7–0.8 Acceptable
0.8–0.9 Good
>0.9 Excellent (but beware of overfitting!)
  • Higher AUC = better probability ranking model
  • ROC shape: Closer to top-left = better

📊 Precision-Recall Curve (Bonus)

Better than ROC when working with highly imbalanced datasets.

from sklearn.metrics import precision_recall_curve
precision, recall, thresholds = precision_recall_curve(y_test, clf.predict_proba(X_test)[:,1])
plt.plot(recall, precision)
  • Visualizes trade-off between precision and recall directly
  • Use when positives are rare (e.g., fraud detection)

✅ Visual Summary

Visual Type Primary Goal When to Use
Confusion Matrix Understand class prediction errors Always
Classification Report Summarize precision/recall/F1 Always
ROC Curve + AUC Measure ranking ability Binary balanced/mid imbalance datasets
Precision-Recall Curve Classify rare-event cases Imbalanced datasets (rare positives)

Would you like a mini notebook that auto-generates all these plots after fitting a model?

  • [[Links]]