Mehmet EmrahKonyaTR
← Back to glossary

Dataset

Veri kümesi

D1

A dataset is a collection of data.

Review status: 2027-02-23

Technical explanation

In an AI workflow, datasets can be designated as training, validation, or testing data with distinct fitting, tuning, and evaluation roles.

Conceptual boundaries

A dataset is data, not the model trained with it; training, validation, and testing data have different stated roles.

Provider-neutral example

A spreadsheet of labelled support tickets can be a training dataset, while a separate held-out set can test the resulting system.

Limitations

A training dataset may not represent the intended-use context or can become stale relative to deployment.

Related concepts

Atomic claims and evidence

  1. 1.1NIST Computer Security Resource Center Glossary, Dataset
    Source
    NIST Computer Security Resource Center Glossary, Dataset
    Source role
    Authoritative source
    Exact locator
    Dataset definition, NIST SP 800-188 entry
    Supported claim
    NIST defines a dataset as a collection of data.
    Last verification
    Review due
    Scope limitation
    The concise definition does not prescribe a file format, size, or use case.
  2. 2.1Regulation (EU) 2024/1689 (Artificial Intelligence Act)
    Source
    Regulation (EU) 2024/1689 (Artificial Intelligence Act)
    Source role
    Authoritative source
    Exact locator
    Article 3(29)-(32)
    Supported claim
    The EU AI Act distinguishes training data used to fit learnable parameters, validation data used to evaluate a trained system and tune its non-learnable parameters and learning process, and testing data used to independently evaluate the system before it is placed on the market or put into service.
    Last verification
    Review due
    Scope limitation
    These are the Act’s defined roles and do not require every technical workflow to use identical dataset splits.
  3. 3.1NIST AI 100-1, Artificial Intelligence Risk Management Framework (AI RMF 1.0)
    Source
    NIST AI 100-1, Artificial Intelligence Risk Management Framework (AI RMF 1.0)
    Source role
    Authoritative source
    Exact locator
    Appendix B, p. 38, data representation and dataset context bullets
    Supported claim
    Training datasets may not represent an AI system’s intended-use context and can become detached from it or stale relative to deployment.
    Last verification
    Review due
    Scope limitation
    This is a risk consideration, not proof that a particular dataset is unsuitable.