Metode za razlaganje nepravičnosti v strojnem učenju

Mlinarič, Nejc

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Metode za razlaganje nepravičnosti v strojnem učenju
Authors:	ID Mlinarič, Nejc (Author) ID Karakatič, Sašo (Mentor) More about this mentor... ID Colakovic, Ivona (Comentor)
Files:	MAG_Mlinaric_Nejc_2026.pdf (2,57 MB) MD5: B6C984219068FF3B720F9AF874A55345
Language:	Slovenian
Work type:	Master's thesis/paper
Typology:	2.09 - Master's Thesis
Organization:	FERI - Faculty of Electrical Engineering and Computer Science
Abstract:	Naloga obravnava razvoj dveh post-hoc metod za razlago nepravičnosti v modelih strojnega učenja z uporabo odločitvenih dreves. Metodi se osredotočata na individualno nepravičnost, pri čemer se lokalne napake agregirajo v pojasnjevalnem drevesu, agregacija preko občutljive spremenljivke pa omogoča vpogled v širše vzorce nepravičnosti. Namen naloge je bil preveriti, ali lahko razviti metodi zaznata različne oblike nepravičnih vzorcev na dveh sintetičnih in enem realnem podatkovnem naboru. Rezultati kažejo, da vsaka metoda zaznava nepravičnost v skladu s svojim konceptualnim pristopom, pri čemer prva metoda izkazuje višje povprečne vrednosti na sintetičnih naborih, druga metoda pa na realnem podatkovnem naboru. Primerjava uteženih in neuteženih rezultatov kaže, da uteževanje prispeva k preglednejši in primerljivejši interpretaciji zaznanih razlik med skupinami.
Keywords:	nepravičnost v strojnem učenju, razložljiva umetna inteligenca (XAI), pravičnost v umetni inteligenci, post-hoc pristop, pojasnjevalni modeli
Place of publishing:	Maribor
Place of performance:	Maribor
Publisher:	[N. Mlinarič]
Year of publishing:	2026
Number of pages:	1 spletni vir (1 datoteka PDF (VII, 57 str.))
PID:	20.500.12556/DKUM-97050
UDC:	004.85(043.2)
COBISS.SI-ID:	272809987
Publication date in DKUM:	03.03.2026
Views:	161
Downloads:	44
Metadata:
Categories:	KTFMB - FERI
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:	CC BY-NC-ND 4.0, Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Link:	http://creativecommons.org/licenses/by-nc-nd/4.0/
Description:	The most restrictive Creative Commons license. This only allows people to download and share the work for no commercial gain and for no other purposes.
Licensing start date:	13.02.2026

Secondary language

Language:	English
Title:	Methods for explaining unfairness in machine learning
Abstract:	This work addresses the development of two post-hoc methods for explaining unfairness in machine learning models using decision trees. The proposed two methods focus on individual unfairness, where locally defined errors are aggregated in a surrogate decision tree, while aggregation through a sensitive attribute enables the identification of broader patterns of unfairness. The aim of the work was to examine whether the developed methods can detect different forms of unfair patterns in two synthetic datasets and one real dataset. The results show that each method detects unfairness according to its conceptual approach. The first method is showing higher average values on synthetic datasets, while the second method on the real dataset. A comparison of weighted and unweighted results shows that weighting contributes to a clearer and more comparable interpretation of perceived differences between groups.
Keywords:	unfairness in machine learning, explainable artificial intelligence (XAI), fairness in artificial intelligence, post-hoc approach, surrogate models

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back