Discriminatory Power#
A model’s ability to separate positive from negative cases.
Important
✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.
What it is#
Discriminatory power is a model’s (or test’s) ability to distinguish correctly between two groups — usually positives versus negatives: good versus bad credit, disease versus healthy, responder versus not. The plain question it answers: how well can the model separate those who will do X from those who won’t?
Where it applies#
A classifier with high discriminatory power assigns higher scores to positives than negatives. A credit scorecard is judged on how cleanly it separates defaulters from non-defaulters; a diagnostic test on how well it separates the sick from the healthy.
How it’s measured#
Several metrics. AUC is the probability a random positive outscores a random negative — 0.5 is random, 1.0 is perfect. The KS statistic is the maximum gap between the cumulative score distributions of positives and negatives (0.4-0.6 is strong in credit risk). The Gini coefficient rescales AUC, \(\text{Gini} = 2 \times \text{AUC} - 1\). Lift and CAP curves give a visual read.
An example, and why it matters#
If good borrowers average a score of 0.8 and bad ones 0.3, with AUC 0.85 and KS 0.45, the model has strong discriminatory power; AUC 0.55, KS 0.08 is near-random. It drives better targeting and lending decisions, is a regulatory reporting requirement in finance, and reduces false positives and negatives for fairer outcomes.
Theme: Classification & Averaging Metrics · All terminology
Hint
Mind map — connected ideas
Gini Coefficient · Multiclass AUROC · KS Statistic (Kolmogorov–Smirnov Statistic) · Demographic Parity (Statistical Parity) · Recalibration
Hint
More in Classification & Averaging Metrics
Accuracy · AUC (Area Under the Curve) · Average Precision (AP) · Binary Classification · Classification Probability · F1-score · Gini Coefficient · Harmonic Mean · Log Loss (also called Logarithmic Loss or Cross-Entropy Loss) · Macro AUC · Macro AUROC (Macro-Averaged AUROC) · Macro Averaging · Macro F1 · Macro Precision
See also
Source article Adapted (context, re-expressed) in our own words from: Discriminatory Power (insightful-data-lab.com).