[lift-curve] Model Lift Chart

[MarkusNeusinger]

Plot Type

lift-curve

Description

Lift chart showing model performance compared to random baseline, commonly used for evaluating marketing and churn prediction models.

Applications

• Marketing campaign targeting
• Customer churn prediction
• Fraud detection evaluation
• Direct mail optimization

Data Requirements

• Predicted probabilities
• Actual binary outcomes
• Decile or percentile groupings

Notes

• Show cumulative lift vs baseline
• Include random model reference line
• X-axis: percentage of population
• Y-axis: lift ratio or cumulative response rate

[claude]

✅ Spec ID: lift-curve

Rationale: "Lift curve" is the established data science term for this model evaluation chart. No similar spec exists - related ML evaluation plots (roc-curve, precision-recall, calibration-curve) serve different purposes.

Add label approved to begin generation.

[github-actions]

✅ Specification Ready: lift-curve

PR: #2382

---

specification.md

lift-curve: Model Lift Chart

Description

A lift curve visualizes how much better a predictive model performs compared to random selection, showing the cumulative lift ratio as you target increasing percentages of the population. It answers the question: "If I target the top X% of predictions, how many times more responders will I capture than random targeting?" This plot is essential for evaluating and comparing classification models in scenarios where targeting efficiency matters.

Applications

• Marketing campaign targeting: Determine how many customers to contact for maximum response rate improvement
• Customer churn prediction: Identify high-risk customers to prioritize for retention efforts
• Fraud detection: Evaluate how effectively a model concentrates fraudulent cases in top-ranked predictions
• Direct mail optimization: Calculate the optimal mailing list size for cost-effective campaigns

Data

• y_true (binary) - Actual outcomes (0 or 1), such as responded/not-responded or churned/retained
• y_score (float) - Predicted probabilities or scores from the model, higher values indicating higher likelihood
• Size: 100-10000 samples recommended for smooth curves
• Example: Customer response data with model probability scores and actual response flags

Notes

• X-axis shows percentage of population targeted (0-100%)
• Y-axis shows cumulative lift ratio (model response rate / baseline response rate)
• Always include a horizontal reference line at y=1 representing random selection (no lift)
• Higher lift at lower percentages indicates better model discrimination
• Curve should start high and gradually approach 1 as percentage increases
• Consider showing decile markers or actual values at key percentiles

---

Next: Add approved label to merge this specification to main.

---

🤖 spec-create workflow

[github-actions]

🚀 Specification Merged!

Specification: lift-curve
PR: #2382 (merged)

The specification is now in main and synced to PostgreSQL.

---

Next: Add generate:{library} labels to start generating implementations.

Available libraries: matplotlib, seaborn, plotly, bokeh, altair, plotnine, pygal, highcharts, letsplot

Or use workflow dispatch for bulk operations.

---

🤖 spec-create workflow

[MarkusNeusinger]

Implementation complete: 9/9 libraries