SIST EN ISO/IEC 5259-2:2025

SLOVENSKI STANDARD
01-julij-2025
Umetna inteligenca - Kakovost podatkov za analizo in strojno učenje - 2. del:
Merjenja kakovosti podatkov (ISO/IEC 5259-2:2024)
Artificial intelligence - Data quality for analytics and machine learning (ML) - Part 2: Data
quality measures (ISO/IEC 5259-2:2024)
Künstliche Intelligenz - Datenqualität für Analytik und maschinelles Lernen (ML) - Teil 2:
Datenqualitätsmaßnahmen (ISO/IEC 5259-2:2024)
Intelligence artificielle - Qualité des données pour les analyses de données et
l’apprentissage automatique - Partie 2: Mesure de la qualité des données (ISO/IEC 5259
-2:2024)
Ta slovenski standard je istoveten z: EN ISO/IEC 5259-2:2025
ICS:
35.020 Informacijska tehnika in Information technology (IT) in
tehnologija na splošno general
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.

EUROPEAN STANDARD EN ISO/IEC 5259-2

NORME EUROPÉENNE
EUROPÄISCHE NORM
May 2025
ICS 35.020
English version
Artificial intelligence - Data quality for analytics and
machine learning (ML) - Part 2: Data quality measures
(ISO/IEC 5259-2:2024)
Intelligence artificielle - Qualité des données pour les Künstliche Intelligenz - Datenqualität für Analytik und
analyses de données et l'apprentissage automatique - maschinelles Lernen (ML) - Teil 2:
Partie 2: Mesure de la qualité des données (ISO/IEC Datenqualitätsmaßnahmen (ISO/IEC 5259-2:2024)
5259-2:2024)
This European Standard was approved by CEN on 18 May 2025.

CEN and CENELEC members are bound to comply with the CEN/CENELEC Internal Regulations which stipulate the conditions for
giving this European Standard the status of a national standard without any alteration. Up-to-date lists and bibliographical
references concerning such national standards may be obtained on application to the CEN-CENELEC Management Centre or to
any CEN and CENELEC member.
This European Standard exists in three official versions (English, French, German). A version in any other language made by
translation under the responsibility of a CEN and CENELEC member into its own language and notified to the CEN-CENELEC
Management Centre has the same status as the official versions.

CEN and CENELEC members are the national standards bodies and national electrotechnical committees of Austria, Belgium,
Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy,
Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Republic of North Macedonia, Romania, Serbia,
Slovakia, Slovenia, Spain, Sweden, Switzerland, Türkiye and United Kingdom.

CEN-CENELEC Management Centre:
Rue de la Science 23, B-1040 Brussels
© 2025 CEN/CENELEC All rights of exploitation in any form and by any means
Ref. No. EN ISO/IEC 5259-2:2025 E
reserved worldwide for CEN national Members and for
CENELEC Members.
Contents Page
European foreword . 3

European foreword
The text of ISO/IEC 5259-2:2024 has been prepared by Technical Committee ISO/IEC JTC 1
"Information technology” of the International Organization for Standardization (ISO) and has been
taken over as EN ISO/IEC 5259-2:2025 by Technical Committee CEN-CENELEC/ JTC 21 “Artificial
Intelligence” the secretariat of which is held by DS.
This European Standard shall be given the status of a national standard, either by publication of an
identical text or by endorsement, at the latest by November 2025, and conflicting national standards
shall be withdrawn at the latest by November 2025.
Attention is drawn to the possibility that some of the elements of this document may be the subject of
patent rights. CEN-CENELEC shall not be held responsible for identifying any or all such patent rights.
Any feedback and questions on this document should be directed to the users’ national standards body.
A complete listing of these bodies can be found on the CEN and CENELEC websites.
According to the CEN-CENELEC Internal Regulations, the national standards organizations of the
following countries are bound to implement this European Standard: Austria, Belgium, Bulgaria,
Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland,
Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Republic of
North Macedonia, Romania, Serbia, Slovakia, Slovenia, Spain, Sweden, Switzerland, Türkiye and the
United Kingdom.
Endorsement notice
The text of ISO/IEC 5259-2:2024 has been approved by CEN-CENELEC as EN ISO/IEC 5259-2:2025
without any modification.
International
Standard
ISO/IEC 5259-2
First edition
Artificial intelligence — Data
2024-11
quality for analytics and machine
learning (ML) —
Part 2:
Data quality measures
Intelligence artificielle — Qualité des données pour les analyses
de données et l’apprentissage automatique —
Partie 2: Mesure de la qualité des données
Reference number
ISO/IEC 5259-2:2024(en) © ISO/IEC 2024

ISO/IEC 5259-2:2024(en)
© ISO/IEC 2024
All rights reserved. Unless otherwise specified, or required in the context of its implementation, no part of this publication may
be reproduced or utilized otherwise in any form or by any means, electronic or mechanical, including photocopying, or posting on
the internet or an intranet, without prior written permission. Permission can be requested from either ISO at the address below
or ISO’s member body in the country of the requester.
ISO copyright office
CP 401 • Ch. de Blandonnet 8
CH-1214 Vernier, Geneva
Phone: +41 22 749 01 11
Email: copyright@iso.org
Website: www.iso.org
Published in Switzerland
© ISO/IEC 2024 – All rights reserved
ii
ISO/IEC 5259-2:2024(en)
Contents Page
Foreword .v
Introduction .vi
1 Scope .1
2 Normative references .1
3 Terms and definitions .1
4 Symbols and abbreviated terms. 5
5 Data quality components and data quality models for analytics and machine learning . 5
5.1 Data quality components in data life cycle .5
5.2 Data quality model .6
6 Data quality characteristics and quality measures .8
6.1 General .8
6.2 Inherent data quality characteristics .9
6.2.1 Accuracy .9
6.2.2 Completeness .10
6.2.3 Consistency . 12
6.2.4 Credibility . 13
6.2.5 Currentness .14
6.3 Inherent and system-dependent data quality characteristics . 15
6.3.1 Accessibility . 15
6.3.2 Compliance . 15
6.3.3 Efficiency .16
6.3.4 Precision .16
6.3.5 Traceability .17
6.3.6 Understandability .17
6.4 System-dependent data quality characteristics .18
6.4.1 Availability .18
6.4.2 Portability .18
6.4.3 Recoverability .19
6.5 Additional data quality characteristics .19
6.5.1 Auditability.19
6.5.2 Balance . 20
6.5.3 Diversity . . 22
6.5.4 Effectiveness . 23
6.5.5 Identifiability .24
6.5.6 Relevance . 25
6.5.7 Representativeness . 25
6.5.8 Similarity . . . 26
6.5.9 Timeliness .27
7 Implementing a data quality model and data quality measures for an analytics or ML
task .28
8 Data quality reporting .28
8.1 Data quality reporting framework . 28
8.2 Data quality measure information . 29
8.3 Guidance to organizations . 29
Annex A (informative) Design and document of a measurement function .30
Annex B (informative) UML model of data quality measure framework .32
Annex C (informative) Overview of data quality characteristics .33
Annex D (informative) Alternative groups of data quality characteristics .35

© ISO/IEC 2024 – All rights reserved
iii
ISO/IEC 5259-2:2024(en)
Annex E (informative) Comparison between data quality characteristics of ISO/IEC 25012 and
ISO/IEC 5259-2 .36
Bibliography .37

© ISO/IEC 2024 – All rights reserved
iv
ISO/IEC 5259-2:2024(en)
Foreword
ISO (the International Organization for Standardization) and IEC (the International Electrotechnical
Commission) form the specialized system for worldwide standardization. National bodies that are
members of ISO or IEC participate in the development of International Standards through technical
committees established by the respective organization to deal with particular fields of technical activity.
ISO and IEC technical committees collaborate in fields of mutual interest. Other international organizations,
governmental and non-governmental, in liaison with ISO and IEC, also take part in the work.
The procedures used to develop this document and those intended for its further maintenance are described
in the ISO/IEC Directives, Part 1. In particular, the different approval criteria needed for the different types
of document should be noted. This document was drafted in accordance with the editorial rules of the ISO/
IEC Directives, Part 2 (see www.iso.org/directives or www.iec.ch/members_experts/refdocs).
ISO and IEC draw attention to the possibility that the implementation of this document may involve the
use of (a) patent(s). ISO and IEC take no position concerning the evidence, validity or applicability of any
claimed patent rights in respect thereof. As of the date of publication of this document, ISO and IEC had not
received notice of (a) patent(s) which may be required to implement this document. However, implementers
are cautioned that this may not represent the latest information, which may be obtained from the patent
database available at www.iso.org/patents and https://patents.iec.ch. ISO and IEC shall not be held
responsible for identifying any or all such patent rights.
Any trade name used in this document is information given for the convenience of users and does not
constitute an endorsement.
For an explanation of the voluntary nature of standards, the meaning of ISO specific terms and expressions
related to conformity assessment, as well as information about ISO's adherence to the World Trade
Organization (WTO) principles in the Technical Barriers to Trade (TBT) see www.iso.org/iso/foreword.html.
In the IEC, see www.iec.ch/understanding-standards.
This document was prepared by Joint Technical Committee ISO/IEC JTC 1, Information technology,
Subcommittee SC 42, Artificial Intelligence.
A list of all parts in the ISO/IEC 5259 series can be found on the ISO and IEC websites.
Any feedback or questions on this document should be directed to the user’s national standards
body. A complete listing of these bodies can be found at www.iso.org/members.html and
www.iec.ch/national-committees.

© ISO/IEC 2024 – All rights reserved
v
ISO/IEC 5259-2:2024(en)
Introduction
Data-supported decision-making brings new challenges to data quality management in data analytics and
artificial intelligence (AI) based on machine learning (ML). Issues in data quality, such as incomplete, false
or outdated data, can adversely affect analytics and ML processes and outcomes. Data from various sources,
including structured data (e.g. relational databases) and unstructured data (e.g. documents, images,
audios), can be directly consumed into the data life cycle for analytics and ML model development. Data
are transformed in each stage of the data life cycle of analytics and ML. A holistic standardized approach to
control, produce and deliver sufficient high-quality data is necessary for data analytics and ML models to be
safe, reliable and interoperable. To develop credible data quality management for analytics and ML, intrinsic
data quality International Standards, including concepts and use cases, characteristics and measurements,
management requirements, and process framework, can be considered.
This document is a part of the ISO/IEC 5259 series. This document builds upon the ISO 8000 series,
ISO/IEC 25012 and ISO/IEC 25024. The purpose of this document is to describe a data quality model through
the definition of data quality characteristics and data quality measures based on ISO/IEC 25012 and
ISO/IEC 25024. Data quality models can be extended or modified according to this document.

© ISO/IEC 2024 – All rights reserved
vi
International Standard ISO/IEC 5259-2:2024(en)
Artificial intelligence — Data quality for analytics and
machine learning (ML) —
Part 2:
Data quality measures
1 Scope
This document specifies a data quality model, data quality measures and guidance on reporting data quality
in the context of analytics and machine learning (ML).
This document is applicable to all types of organizations who want to achieve their data quality objectives.
2 Normative references
The following documents are referred to in the text in such a way that some or all of their content constitutes
requirements of this document. For dated references, only the edition cited applies. For undated references,
the latest edition of the referenced document (including any amendments) applies.
ISO/IEC 5259-1, Artificial intelligence — Data quality for analytics and machine learning (ML) — Part 1:
Overview, terminology, and examples
ISO/IEC 25024, Systems and software engineering — Systems and software Quality Requirements and
Evaluation (SQuaRE) — Measurement of data quality
ISO/IEC 22989, Information technology — Artificial intelligence — Artificial intelligence concepts and
terminology
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO/IEC 5259-1, ISO/IEC 22989 and
the following apply.
ISO and IEC maintain terminology databases for use in standardization at the following addresses:
— ISO Online browsing platform: available at https:// www .iso .org/ obp
— IEC Electropedia: available at https:// www .electropedia .org/
3.1
data
re-interpretable representation of information in a formalized manner suitable for communication,
interpretation, or processing
Note 1 to entry: Data can be processed by humans or by automatic means.
[SOURCE: ISO/IEC 2382:2015, 2121272]

© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
3.2
data frame
set of data records represented by a specific domain or purpose, with a shared structure of data items
Note 1 to entry: A data frame is two-dimensional, like a table with rows and columns. The term is specifically used in
analytics and ML, e.g. in the R language, while other languages use “data set” to mean the same thing. In this document,
“dataset” has a more generic meaning.
3.3
data type
categorization of an abstract set of possible values, characteristics, and set of operations for an attribute
Note 1 to entry: Examples of data types are character strings, texts, dates, numbers, images and sounds.
[SOURCE: ISO/IEC 25024:2015, 4.16]
3.4
data value
content of data item
Note 1 to entry: In ISO/IEC 25012:2008, 5.1.1, it is specified that from the inherent point of view, data quality refers to
data itself such as data domain values and possible restrictions.
Note 2 to entry: Number or category assigned to an attribute of a target entity by making a measurement.
[SOURCE: ISO/IEC 25024:2015, 4.17]
3.5
empty data item
data item whose data value (3.4) has no value, i.e. Null or None
Note 1 to entry: This definition in general signifies non-existence of a data value (i.e. Null or None). A data item with
string data type can be an empty data item by using either the empty string or Null. However, there is an exception for
some application a string can be empty (e.g. “”) but not Null and hence not imply an empty data item.
3.6
entity
concrete or abstract thing in the domain under consideration
[SOURCE: ISO 8000-2:2022, 3.3.3]
3.7
raw data
data in its originally acquired, direct form from its source before subsequent processing
[SOURCE: ISO 5127:2017, 3.1.10.04]
3.8
target data
data (3.1) used in an analytics or ML task whose quality is measured
3.9
target population
population of interest in the analytics or ML project to which inferences are to be made
3.10
data quality subject
entity (3.6) affected by data quality

© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
3.11
quality measure element
measure defined in terms of a property and the measurement method for quantifying it, including optionally
the transformation by a mathematical function
[SOURCE: ISO/IEC 25024:2015, 4.32]
3.12
quantity
property of a phenomenon, body, or substance, where the property has a magnitude that can be expressed
as a number and a reference
[SOURCE: ISO/IEC Guide 99:2007, 1.1, modified — Notes to entry deleted.]
3.13
quantity value
number and reference together expressing magnitude of a quantity (3.12)
[SOURCE: ISO/IEC Guide 99:2007, 1.9, modified — Examples deleted.]
3.14
measurement function
algorithm or calculation performed to combine one or more quality measure elements (3.11)
[SOURCE: ISO/IEC 25021:2012, 4.7, modified — Definition revised.]
3.15
measurement result
result of measurement
set of quantity values (3.13) being attributed to a measurand together with any other available relevant
information
[SOURCE: ISO/IEC Guide 99:2007, 2.9, modified — Notes to entry deleted.]
3.16
measure
variable to which a value is assigned as the result of measurement
Note 1 to entry: The plural form “measures” is used to refer collectively to base measures, derived measures and
indicators.
[SOURCE: ISO/IEC/IEEE 15939:2017, 3.15]
3.17
measure
make a measurement
[SOURCE: ISO/IEC 25000:2014, 4.19]
3.18
bounding box
rectangular region enclosing annotated object
Note 1 to entry: The major and minor axes of the rectangle are parallel to the edges of the images. For rotated boxes,
the polygon annotation is to be used.
[SOURCE: ISO/IEC 30137-4:2021, 3.3]
3.19
cluster
automatically induced category of elements that are part of the dataset and that share common attributes
Note 1 to entry: Clusters do not necessarily have a name.

© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
[SOURCE: ISO/IEC 23053:2022, 3.3.2]
3.20
clustering algorithm
algorithm which groups clusters (3.19) from input data
Note 1 to entry: Examples of clustering algorithms include centroid-based clustering, density-based clustering,
distribution-based clustering, hierarchical clustering and graph-based clustering.
3.21
overfitting
creating a model which fits the training data too precisely and fails to generalize on
new data
Note 1 to entry: Overfitting can occur because the trained model has learned from non-essential features in the
training data (i.e. features that do not generalize to useful outputs), excessive noise in the training data (e.g. excessive
number of outliers), a significant mismatch between training data and production data distributions or because the
model is too complex for the training data.
Note 2 to entry: Overfitting can be identified when there is a significant difference between errors measured on
training data and on separate test and validation data. The performance of overfitted models is especially impacted
when there is a significant mismatch between training data and production data.
[SOURCE: ISO/IEC 23053:2022, 3.1.4]
3.22
fidelity
degree to which a model or simulation reproduces the state and behaviour of a real-world object or the
perception of a real-world object, feature, condition, or chosen standard in a measurable or perceivable manner
[SOURCE: ISO 16781:2021, 3.1.4]
3.23
maintainability
ability of a functional unit, under given conditions of use, to be retained in, or restored to, a state in which it
can perform a required function when maintenance is performed under given conditions and using stated
procedures and resources
Note 1 to entry: The term used in IEV 191-02-07 is “maintainability performance” and the definition is the same.
Note 2 to entry: maintainability: term and definition standardized by ISO/IEC [ISO/IEC 2382-14:1997].
Note 3 to entry: 14.01.06 (2382)
[SOURCE: ISO/IEC 2382:2015, 2123027]
3.24
reliability
consistency with which an assessment measures
EXAMPLE An assessment will have low reliability if two assessment forms are of unequal difficulty or coverage
or if there are errors in the scoring procedures or in the reporting of scores.
[SOURCE: ISO/IEC 23988:2007, 3.21]
3.25
validity
extent to which an assessment achieves its aim by measuring what it is supposed to measure and producing
results which can be used for their intended purpose
Note 1 to entry: An assessment has low validity if the results are unduly influenced by skills which are irrelevant to
the stated aims of the assessment.
[SOURCE: ISO/IEC 23988:2007, 3.25]

© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
4 Symbols and abbreviated terms
AI artificial intelligence
CSV comma separated values
HDF hierarchical data format
JSON JavaScript object notation
ML machine learning
IP internet protocol
PII personally identifiable information
QM quality measure
UML unified modelling language
5 Data quality components and data quality models for analytics and machine
learning
5.1 Data quality components in data life cycle
Figure 1 shows data quality components aligned with the data life cycle model shown in ISO/IEC 5259-1:2024,
Figure 3, which can support data quality management processes. ISO/IEC 5259-1 defines a data quality
model as a defined set of data quality characteristics. The data quality characteristic provides a framework
for data quality requirements, implementation and evaluation methods. Data quality measures are variables
assigned to which values are the results of measurements of data quality characteristics. Data quality
measures are used to assess whether the data meet data quality requirements. Data quality measures can
also be used to monitor and report data quality.
Target data are the data subject to data quality measurements. Target data can be raw data or data that has
undergone one or more processes or transformations. Target data for measuring quality can be training,
testing, validation, production and output data in the context of the use of analysis and ML (as described
[1]
in ISO/IEC 23053). Target data can be formed as either data items or datasets. A data item consists of an
item name, data value and data type representing a domain of values (e.g. character strings, texts, dates,
numbers, images, sounds). A dataset can be classified into three forms:
— a collection of data items;
— a collection of data records;
— a collection of data frames.
The target data can be unlabelled or labelled depending on the association with data labels in the use of
analytics or ML task.
NOTE This document makes no distinction between data structures, such as structured data, semi-structured
data and unstructured data, or data roles, such as master data, transaction data and reference data.
Data quality reports are documents that express data quality requirements, the data quality model of data
quality characteristics, data quality measures, the results of data quality measurements and an assessment
of whether the data meet data quality requirements.

© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
Key
Stage where data are processed
Data quality component
Primary development pathway
Dependency
Feedback pathway
Figure 1 — Data quality components in data life cycle for analytics and ML
5.2 Data quality model
The data quality model provides a framework for specifying data quality requirements and evaluating data
quality. In practice, a data quality model brings together data quality subjects, data quality characteristics
and data quality requirements, for the context of the use of the data. The organization can specify data
quality models by selecting data quality characteristics and measures to achieve target quality requirements
for target data. Figure 2 provides a UML diagram of the relationships between the components of the data
quality model.
A data usage scope describes how and where the data can be used in an analytics or ML task and how it fits
into an AI system.
EXAMPLE The data can be used to train a deep neural network ML model to predict product sales based on the
features of a marketing strategy. The model can be trained and deployed using cloud services.
A data quality subject represents an entity affected by data quality. A data quality characteristic is a category
of data quality attributes that bear on data quality (e.g. accuracy, completeness, precision). A data quality
requirement describes properties or attributes of the data along with acceptance criteria relative to the data
usage scope. Acceptance criteria can be quantitative or qualitative.

© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
Figure 2 — Data quality model
When one quality characteristic affects another, trade-offs can be made by evaluating each requirement
regarding importance and impact. In addition, it is crucial to balance the cost of data quality management
with the priority of data quality requirements in determining how data quality characteristics and measures
are incorporated into the data quality model. The organization can select the data quality characteristics
and measures that correspond to their needs and requirements. Data quality should be assessed by
comparing the results of selected data quality measures against established targets as established by data
[2]
requirements. Any failures to achieve data quality requirements should be mitigated. ISO/IEC 5259-3
describes the requirements and recommendations of a data quality management system to be applied by the
organization.
[3] [4]
ISO 8000-8 and ISO/IEC 25012 describe data quality models. ISO 8000-8 defines three data quality
characteristics as being syntactic (format), semantic (meaning), and pragmatic (usefulness) to support
industrial data generally as a product of business and manufacturing processes. ISO/IEC 25012 defines a
general data quality model for data retained in a structured format within a computer system as a part
of a software product. ISO/IEC 25012 takes into account all data types (e.g. characters, strings, texts,
dates, numbers, images, sounds). ISO/IEC 25012 provides fifteen data quality characteristics: accuracy,
completeness, consistency, credibility, currentness, accessibility, compliance, confidentiality, efficiency,
precision, traceability, understandability, availability, portability and recoverability.
[5]
The ISO 8000 series addresses various aspects of data quality such as data governance, data quality
[6]
management (including processing) and maturity assessment. The ISO/IEC 25000 series addresses
product (software, systems, data, services) quality requirements and evaluation. This document describes
how the data quality characteristics of ISO/IEC 25012 can be applied to a data quality model for analytics
and ML. Furthermore, this document defines additional characteristics that can contribute to higher-quality
ML models and applications, as shown in Figure 3. Organizations should use the data quality characteristics
and data quality measures described in this document whenever possible. However, the data quality
characteristics in this document cannot comprehensively cover aspects that support all organizations’
needs regarding data quality. Organizations may design their own data quality model by extending the data
quality characteristics and data quality measures to fit their data requirements.
NOTE 1 See Annex A for information on designing and documenting measurement functions.
NOTE 2 See Annex E for a comparison between the data quality characteristics in ISO/IEC 25012 and those in this
document.
© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
Figure 3 — Data quality characteristics for analytics and ML
6 Data quality characteristics and quality measures
6.1 General
Data quality characteristics and measures are used to specify and verify data quality requirements for
identified attributes for target data. Each data quality characteristic is associated with one or more data
quality measures for quantification. A data quality measure is a variable to which a value is assigned as the
result of a measurement function. The data quality measures in this document are selected based on the
context of use of analytics and ML.
NOTE 1 Annex B shows a framework for providing common vocabularies and relationships between the components
of data quality measures.
NOTE 2 Annex C and Annex D show how quality measures are grouped from different perspectives.
In the context of analytics and ML, the overall quality of a training dataset, a validation dataset or a test
dataset can be just as important as the quality of the individual data values in the dataset. Even though every
data value in a dataset is accurate, a dataset that does not correctly reflect the underlying distribution of
data can cause an incorrect analysis result or the creation of an ML model that does not meet requirements.
The organization should document the target data for each data quality measure.
NOTE 3 Characteristics for statistical measures (e.g. accessibility by authorized users, accuracy, consistency,
[7]
currentness, understandability, relevance, timeliness) as defined by institutions such as the United Nations
Statistics Division (UNSD) and European Statistics (EUROSTAT) can also be used to assess whether the quality of a
dataset meets requirements.
The data quality measures and measurement functions in this document should be used when appropriate.
Refer to Annex A in cases where the user of this document needs to create a new, bespoke data quality
measure and data quality measurement function. Any quality measure, when using modified or newly
defined, shall select data quality characteristics defined in this document and shall provide the rationale for
changes in accordance with ISO/IEC 25024:2015, Clause 2.

© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
6.2 Inherent data quality characteristics
6.2.1 Accuracy
6.2.1.1 General
Accuracy of a dataset is the degree to which data items in the data set have the correct data values or correct
data labels. ISO/IEC 25012 describes accuracy as the degree to which data values have attributes that
correctly represent the true value of the intended attributes. ISO/IEC 25012 further describes accuracy in
terms of:
— syntactic accuracy which considers the closeness of the data values to a set of syntactically correct data
values in a relevant domain;
— semantic accuracy which considers the closeness of the data values to a set of semantically correct data
values in a relevant domain.
A data item is syntactically correct if its data value is the same type as its explicit data type and semantically
correct if its data value has an expected value corresponding to the ML task. ML models are mathematical
constructs, which means that low syntactic or semantic accuracy of the data values in training, validation,
testing or production datasets can cause the model itself to be incorrect or the inferences made by the model
to be incorrect.
For a supervised learning classification system, the correctness of the label sequence contents can affect
the inference accuracy of a trained model. Factors that should be considered for measuring the accuracy of
labelling include:
— correctness of label values;
— correctness of labelled tags;
— correctness of label sequence contents.
EXAMPLE 1
If the phrase “lazy dog” is entered as “lzy dg” an ML-based natural language understanding system can fail to correctly
interpret the phrase.
EXAMPLE 2
If the number 100 is entered as 1000 in training data, a regression model can fail to correctly calculate the weight of
the related feature and if the entry was made in the production data, inferences can be incorrect.
6.2.1.2 QMs for accuracy
Table 1 provides data quality measures for accuracy in a specific context of use of analytics and ML.

© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
Table 1 — Accuracy measures
ID Name Description Measurement function
Acc-ML-1 Syntactic data See ISO/IEC 25024:2015, Table 1 See ISO/IEC 25024:2015, Table 1
accuracy
Acc-ML-2 Semantic data See ISO/IEC 25024:2015, Table 1 See ISO/IEC 25024:2015, Table 1
accuracy
Acc-ML-3 Data accuracy See ISO/IEC 25024:2015, Table 1 See ISO/IEC 25024:2015, Table 1
assurance
Acc-ML-4 Risk of dataset See ISO/IEC 25024:2015, Table 1 See ISO/IEC 25024:2015, Table 1
inaccuracy
Acc-ML-5 Data model accu- See ISO/IEC 25024:2015, Table 1 See ISO/IEC 25024:2015, Table 1
racy
Acc-ML-6 Data accuracy See ISO/IEC 25024:2015, Table 1 See ISO/IEC 25024:2015, Table 1
range
Acc-ML-7 Data label accuracy Does data label correctly assign to
A
each element in the dataset?
B
where
A is the number of data labels that pro-
vide the appropriate required informa-
tion;
B is the number of data labels defined in
the dataset.
6.2.2 Completeness
6.2.2.1 General
ISO/IEC 25012 describes completeness in terms of data having values for all expected attributes and entity
instances. In some cases, ML algorithms can fail when they encounter one or more empty data items in
training, validation or testing datasets. Additionally, trained ML models can also fail when production data
contains null data values.
Measures for completeness can help ML practitioners meet their data requirements and can indicate
[8]
whether additional imputation steps should be taken as described in ISO/IEC 5259-4.
The completeness characteristic of the labelled data in a dataset is relative. In different scenarios, the
meaning of completeness can be different and should be considered with a specific usage scope. Factors that
should be considered for measuring the completeness of a dataset include:
— The completeness of a dataset being used for an ML-based image classification should check the unlabelled
samples in a dataset, which cannot be directly used in supervised ML.
— The completeness of a dataset being used for an ML-based object detection should check the incompleteness
of labelled bounding boxes on objects.
In particular, it is common in real life that a sample has multiple objects in various categories since it is
difficult to capture a scene with a single isolated object taking the entire view space. In this case, to
measure the completeness of the dataset for an ML-based image recognition, the following factors should be
considered:
— there exists any target object in a sample;
— all target objects are categorized;
— all target objects detected are labelled with bounding boxes or other methods.
EXAMPLE 1
© ISO/IEC 2024 – All rights reserved
ISO/IEC 5259-2:2024(en)
A completeness measure for a dataset indicates that the dataset is missing more than half of the data values for the zip
...