PLAN RADA SEMINARA ZA MART 2025. GODINE
Predavanja na seminaru mogu se pratiti na daljinu preko linka
https://miteam.mi.sanu.ac.rs/asset/wnz6oyxsQsy29LfJA.
Registracija za on-line praćenje predavanja na Seminaru je na linku
https://miteam.mi.sanu.ac.rs/asset/xzGqvSp7aWbg8WpYX.
Neulogovani korisnici mogu pratiti prenos predavanja na ovom linku (ali ne mogu postavljati pitanja osim putem chata):
https://miteam.mi.sanu.ac.rs/call/wnz6oyxsQsy29LfJA/MjQ__eH607WeAL9X7IFtUI98xdQQgVkp-ljiEKPPfXr.
Utorak, 04.03.2025. u 14:15, Knez Mihailova 36, sala 301f i
Online
Tijana Matejić, Akademija strukovnih studija Šumadija
PRIMENA VEŠTAČKIH NEURONSKIH MREŽA I INDIKATORA INTERNET PRISUSTVA U PREDVIĐANJU RIZIKA OD BANKROTSTVA HOTELIJERSKIH PREDUZEĆA
U radu su predstavljeni modeli dubokih Veštačkih neuronskih mreža (VNM) za ordinalnu klasifikaciju, koji su korišćeni za procenu kategorija hotelijerskih preduzeća u pogledu njihovog rizika od bankrotstva u periodu od 2020. do 2024. godine koje odgovaraju kategorijama dobijenim pomoću Altmanovog Z"-Skor modela za tržišta u razvoju i Zmijewski modela. U istraživanju je primenom heurističke metode koja se bazira na Analizi glavnih komponenti, koeficijenata korelacije i potom klasterizacije putem metode K-srednjih vrednosti, razvijen model internet prisustva preduzeća u 2023. godini, koji obuhvata pet kategoričkih indikatora različitih dimenzija internet prisustva koje opisuju tačke dodira koje preduzeće ima sa korisnicima na internetu i predstavljaju skup ulaznih parametara VNM modela. Radi validacije rezultata VNM modela, u istraživanju su razvijeni i primenjeni i analogni modeli Ordinalne logističke regresije. Pri evaluaciji preciznosti svih razvijenih modela korišćena je sistematična metodologija koja: rešava problem sa nebalansiranim kategorijama rizika od bankrotstva primenom metode pre-uzorkovanja (ADASYN), primenjuje unakrsnu validaciju radi robustnijih procena preciznosti modela i koristi četiri indikatora u evaluaciji preciznosti modela.
Utorak, 11.03.2025. u 14:15, Knez Mihailova 36, sala 301f i
Online
Filip Vidojević, Faculty of Mathematics, University of Belgrade
SKEWED VARIABLE NEIGHBORHOOD DECOMPOSITION SEARCH ALGORITHM FOR THE HYBRID FLOW SHOP SCHEDULING PROBLEM
Hybrid Flow Shop (HFS) scheduling plays a crucial role in industries such as glass, steel, paper, and textiles, where optimizing machine utilization directly influences productivity. This paper introduces a novel approach that integrates Skewed Variable Neighborhood Search (SVNS) to minimize makespan, leveraging its adaptive exploration and strategic job sequencing to enhance scheduling efficiency. A key innovation of our method is a specialized encoding scheme that strategically allows jobs to wait, even when machines are available, to enable better sequencing decisions. This mechanism not only improves solution quality but also permits a natural decomposition of the scheduling problem, leading to more balanced machine workloads. By leveraging the flexibility of VNS alongside this tailored encoding, our approach significantly enhances scheduling efficiency. Experimental results confirm its effectiveness, outperforming existing methods in makespan reduction.
This research was conducted in collaboration with Dušan Džamić, Andrijana Džamić, and Miroslav Marić.
Utorak, 18.03.2025. u 14:15, Knez Mihailova 36, sala 301f i
Online
Petar Lukovac i Aleksandar Joksimović, Fakultet organizacionih nauka, Univerzitet u Beogradu
MODELING SMART APICULTURE ECOSYSTEM: AN ONTOLOGY-BASED APPROACH
The introduction of smart devices has proven its value in improving efficiency in ecosystems by introducing remote real-time monitoring and data sharing between stakeholders. This paper presents a model of a smart apiculture ecosystem that uses ontology at its core. The apiculture domain is chosen for its essential role in the global food market and both direct and indirect impact on the majority of agricultural domains. The research goal is to provide an approach to modeling multidomain smart ecosystems. To achieve this, a domain-specific ontology is created, as an extension to SSN ontology. Based on the created ontology, an ecosystem is created, as well as a model for smart apiary IT platform. The used process and the created model bring towards standardization of data in agricultural domain and provides a framework for future development.
Utorak, 25.03.2025. u 14:15, Knez Mihailova 36, sala 301f i
Online
Jelena Jocković, Faculty of Mathematics, University of Belgrade, Serbia
MARKOV CHAIN APPROACH TO THE COUPON COLLECTOR PROBLEM WITH UNIVERSAL COUPON
The classical coupon collector problem has various modifications and generalizations. One group of generalizations is based on the idea of introducing additional coupons, with special purposes, to the set of available coupons. We consider the case when this set consists of standard coupons (that can belong to the collection), a null coupon (which can be drawn, but does not belong to any collection), and an additional universal coupon, that can replace any of the standard coupons. By employing a Markov chain approach, we derive the exact forms of the k-step transition matrix and the fundamental matrix, which we use to obtain the properties of the waiting time until a subcollection, or a full collection is sampled, and some additional characteristics of the collecting process (probability that the coupon collecting procedure ends in a particular way). We also provide numerical examples and explain possible applications of the variant of the coupon collector problem we considered.
The talk is based on the paper J. Jocković, B. Todić,
Markov chain approach to the coupon collector problem with universal coupon, Filomat
38(24) (2024), 8565-8581.
RUKOVODIOCI SEMINARA
MI SANU
Vera Kovačević-Vujčić
Milan Dražić
FON
Zorica Bogdanovic
Marijana Despotovic-Zrakic
IEEE
Bozidar Radenkovic
