ISO IEC 5259-4-2024 PDF

Full title and description

ISO/IEC 5259-4:2024 — Artificial intelligence — Data quality for analytics and machine learning (ML) — Part 4: Data quality process framework. This part of the ISO/IEC 5259 series defines a standardized process framework to manage and assure the quality of data used for analytics and all types of ML (supervised, unsupervised, semi‑supervised and reinforcement learning), with emphasis on data acquisition, composition, preparation, labelling, evaluation, lifecycle management and traceability.

Abstract

Provides organizational-level guidance and a process framework for ensuring data quality across the ML data lifecycle used for training and evaluation. It addresses common organizational approaches to labelling, evaluation and data governance for analytics and ML, without mandating specific tools or platforms. The standard is intended to improve reliability, traceability and reproducibility of ML outcomes by strengthening data-quality processes.

General information

• Status: Published
• Publication date: 15 July 2024
• Publisher: Jointly published under ISO and IEC (ISO/IEC JTC 1/SC 42)
• ICS / categories: 35.020 (Information technology — Artificial intelligence)
• Edition / version: Edition 1.0 (2024)
• Number of pages: 28

General bibliographic details above are based on the ISO and IEC publication records for ISO/IEC 5259-4:2024.

Scope

Establishes a common process framework to manage data quality for analytics and ML training and evaluation data sourced from various origins. The document covers process-level guidance applicable across supervised, unsupervised, semi‑supervised and reinforcement learning as well as analytics, including data acquisition and composition, data preparation, labelling, evaluation, use and lifecycle activities. It is explicitly not a specification of particular services, tools or platforms.

Key topics and requirements

• Data quality process framework spanning acquisition, composition, preparation, labelling, evaluation, use and lifecycle management.
• Guidance on organizational roles, responsibilities and governance for data quality in ML projects.
• Recommendations for data labelling practices and quality assurance for training/evaluation datasets.
• Evaluation and validation processes to measure and document dataset quality, traceability and reproducibility.
• Applicability across ML types (supervised, unsupervised, semi‑supervised, reinforcement) and analytic use cases.
• Emphasis on process-based controls rather than mandating specific tools, platforms or services.

Typical use and users

Intended for organizations and teams that develop, procure or operate ML and analytics systems: ML engineers, data engineers, data scientists, quality managers, AI governance teams, data labelling service providers, auditors and procurement teams. Use cases include establishing repeatable data-quality processes, vendor assessments, internal QA for training datasets, and aligning data workflows with organizational governance.

Related standards

Part of the ISO/IEC 5259 series on AI data quality; related parts include ISO/IEC 5259-1, -2, -3 and -5 which collectively cover principles, definitions and management requirements for AI data quality. It is also aligned with other ISO/IEC JTC 1/SC 42 outputs addressing AI system lifecycle, governance and data management.

Keywords

data quality, machine learning, analytics, data labelling, data lifecycle, data governance, evaluation, traceability, ISO/IEC 5259, SC 42

FAQ

Q: What is this standard?

A: ISO/IEC 5259-4:2024 is Part 4 of the ISO/IEC 5259 series and defines a process framework for assuring data quality used in analytics and machine learning.

Q: What does it cover?

A: It covers process-level guidance across the ML data lifecycle — acquisition, composition, preparation, labelling, evaluation and use — and applies to supervised, unsupervised, semi‑supervised and reinforcement learning. It provides organizational guidance but does not mandate specific tools or services.

Q: Who typically uses it?

A: ML teams, data engineers, data labelling providers, quality and governance teams, auditors and organizations procuring ML datasets or services use this standard to implement and evaluate data‑quality processes.

Q: Is it current or superseded?

A: Current — it was published in July 2024 (ISO/IEC 5259-4:2024) and is the first edition.

Q: Is it part of a series?

A: Yes — it is one part of the ISO/IEC 5259 series on data quality for analytics and ML; other parts address complementary principles and management requirements.

Q: What are the key keywords?

A: Data quality, labelling, ML, analytics, lifecycle, governance, traceability, evaluation, ISO/IEC 5259.