🔬  Coverage

Coverage#

The share of outcomes that fall within predicted intervals.

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 Despite careful review, it may still contain mistakes, omissions, or out-of-date information.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#

Coverage measures how much of the catalog a recommender actually uses — the share of available items it is able to, or chooses to, recommend. A system can be accurate yet only ever surface a handful of popular items, ignoring the rest.

Two flavors#

Prediction coverage is the fraction of items for which the model can make a prediction at all; catalog coverage is the fraction of items that actually appear in the recommendation lists users see. The latter is the usual beyond-accuracy target.

Why it matters#

High coverage means the long tail gets exposure and the catalog isn’t wasted — countering popularity bias. A limitation of plain coverage: it counts an item shown once the same as one shown thousands of times, which is why Gini and entropy refine it to capture how evenly exposure is spread.


Theme: Model Evaluation & Uncertainty  ·  All terminology



See also

Source article Adapted (context, re-expressed) in our own words from: Coverage (insightful-data-lab.com).

Tags: purpose: reference topic: terminology level: intermediate