20.500.12556/DKUM-98427 Primerjalna analiza metod testiranja informacijskih rešitev z inteligentnimi komponentami magistrsko delo Comparative analysis of testing methods for information systems with intelligent components Razvoj inteligentnih sistemov prinaša nove izzive pri testiranju in zagotavljanju njihove kakovosti, saj rezultati temeljijo na verjetnosti namesto na vnaprej določenih izidih. Magistrsko delo obravnava problem testiranja sistemov z inteligentnimi komponentami, pri čemer smo sistematično pregledali obstoječe metode in ogrodja ter izvedli primerjalno analizo metamorfnega, diferencialnega in testiranja na podlagi lastnosti. Cilj naloge je bil ugotoviti, kako se izbrane metode razlikujejo glede na vrsto zaznanih odstopanj in ali njihova kombinacija omogoča širši vpogled v delovanje sistema. V eksperimentu smo vse tri metode implementirali na konvolucijski nevronski mreži, naučeni na javno dostopnem podatkovnem naboru. Rezultate smo primerjali s Kruskal-Wallisovim testom in post-hoc analizo. Ugotovili smo, da metode odkrivajo različne tipe nepravilnosti in imajo različne prednosti in omejitve. Rezultati so pokazali statistično značilne razlike med izbranimi pristopi in potrjujejo njihovo uporabnost pri zagotavljanju kakovosti inteligentnih informacijskih rešitev. The development of intelligent systems introduces new challenges in testing and quality assurance, as results are based on probability rather than predefined outcomes. This master's thesis addresses the problem of testing systems with intelligent components, systematically reviewing existing methods and frameworks and conducting a comparative analysis of metamorphic, differential, and property-based testing. The aim of the thesis was to determine how the selected methods differ in terms of the types of deviations they detect and whether their combination enables a broader insight into system behaviour. In the experiment, all three methods were implemented on a convolutional neural network trained on a publicly available dataset. The results were compared using the Kruskal-Wallis test and post-hoc analysis. We found that the methods detect different types of irregularities and have different strengths and limitations. The results showed statistically significant differences between the selected approaches and confirm their practical applicability in ensuring the quality of intelligent information solutions. testiranje inteligentnih komponent testiranje sistemov z inteligentno komponento inteligentna komponenta metamorfno testiranje testiranje na podlagi lastnosti diferencialno testiranje. intelligent components testing testing systems with intelligent component intelligent component metamorphic testing property based testing differential testing. true true false [H. Zakelšek] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2026-06-11 16:55:33 2026-06-29 14:20:13 2026-07-04 03:09:47 0000-00-00 00:00:00 2026 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (…,… str.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.89(043.2) 283573507 MAG_Zakelsek_Haidi_2026.pdf MAG_Zakelsek_Haidi_2026.pdf 1 D278B8E0CF1B59766F41E194B50EA0B2 2dcff9b45f1f9f0593b29b49a2d24c42e4e08072226829160b86d7ecd1179ac2 b15e0122-6f3f-11f1-973b-00155d580e34 https://dk.um.si/Dokument.php?lang=slv&id=207509 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0