Primerjava generativnih modelov umetne inteligence za generiranje programske kode

magistrsko delo

Jan Ključevšek (Author), Domen Verber (Mentor)

Abstract

Cilj magistrskega dela je bil primerjalno oceniti kakovost programske kode, ki jo generirajo modeli umetne inteligence ChatGPT (4o, o1), Gemini (Flash, Pro) in Microsoft Copilot. Na področju generativne umetne inteligence in kakovosti programske opreme smo z uporabo kvantitativnih metrik in orodij analizirali kodo, generirano za različno zahtevne naloge. Rezultati kažejo, da vsi modeli ustvarjajo sintaktično pravilno kodo, a se razlikujejo predvsem v funkcionalni pravilnosti, kompleksnosti in berljivosti. Plačljivi modeli so bili pravilnejši, a kompleksnejši; brezplačni (Copilot, Gemini Flash) pa enostavnejši in berljivejši. Priporočamo izbiro modela glede na prioritete projekta.

Keywords

generativna umetna inteligenca;generiranje programske kode;kakovost programske kode;metrike kakovosti kode;magistrske naloge;

Data

Language:	Slovenian
Year of publishing:	2025
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[J. Ključevšek]
UDC:	004.8:004.4'415(043.2)
COBISS:	239013891
Views:	0
Downloads:	41
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Comparison of generative artificial intelligence models for code generation
Secondary abstract:	The objective of the master's thesis was to comparatively evaluate the quality of program code generated by the artificial intelligence models ChatGPT (4o, o1), Gemini (Flash, Pro), and Microsoft Copilot. In the field of generative artificial intelligence and software quality, we analyzed code generated for tasks of varying complexity using quantitative metrics and tools. The results show that all models produce syntactically correct code, but differ primarily in functional correctness, complexity, and readability. Paid models were more correct but more complex; free models (Copilot, Gemini Flash) were simpler and more readable. We recommend selecting a model based on project priorities.
Secondary keywords:	generative artificial intelligence;code generation;code quality;code quality metrics;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Informatika in podatkovne tehnologije
Pages:	1 spletni vir (1 datoteka PDF (XIII, 82 str.))
ID:	26401469