Uncertainty and models
Cumulative gains chart
Shows the share of positive outcomes captured as more score-ranked cases are selected.
- Chart
- Model evaluation
- Ranking
What it answers
What share of positives is captured by selecting more cases?
- Shape
- Captured outcomes versus selected share
- What it shows
- Early positive capture and diminishing returns.
Example
The question
How many buyers do we reach by mailing the top-scored customers?
Share of all buyers reached against share of customers contacted (cumulative gains)
Illustrative data. It belongs to no company.
When to use it
Use it when
Evaluating ranking-based prioritisation.
Avoid it when
When score direction or total positives are unclear.
How to read it
Read selected-population share and cumulative positive share.
Note
Specify score direction and tie handling.
The data you need
- Layout
- Long table
- Fields
- CaseID: text; Actual: 0/1; Score: number
- One row is
- One case with its true binary outcome and prediction score. What is the grain?
- Rules
- Define the positive class and score direction. Use held-out predictions for evaluation. Calibration needs probabilities in [0,1].
- Optional
- Weight. ROC/PR alternatives: precomputed Threshold,TP,FP,TN,FN with complete threshold coverage.
- Preparation
- Sort scores, accumulate positives and divide by all positives; compare with fraction of cases selected.
- Units
- X=selected cases/all cases; Y=captured positives/all positives, after ranking byscore.
Worked example
There are 10 positives in 100 cases; top 20 contain 6. Plot(20% selected,60% captured), compared with 20% capture for random ordering onaverage.
| CaseID | Actual | Score |
|---|---|---|
| A | 1 | 0.90 |
| B | 0 | 0.20 |
| C | 1 | 0.65 |
| D | 0 | 0.55 |