Hibridni događaj

Program događaja

srijeda, 27.5.2026 16:00 - 19:00, Bellavista, Grand hotel Adriatic, Opatija Radovi  1. I. Mekterović, L. Brkić, M. Tadijal, T. Vanjak (University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Evolving Programming Education with AI: Opportunities and Challenges of LLM-Integrated Assessment Systems  This paper explores the potential for the use of large language models (LLMs) primarily in the context of automated programming assessment systems (APAS) but also more broadly, in the environment of e-learning systems in higher education. Starting from a review of recent literature and first applications, the paper systematizes the possible roles of artificial intelligence: from automated generation of personalized tasks and provision of formative feedback to intelligent teaching and objective evaluation of complex software solutions. The central part of the paper is dedicated to the specific analysis of content generation for introductory programming courses in higher education (CS1/CS2), with an emphasis on automated synthesis of problem tasks, associated reference solutions and exhaustive testing examples. The pilot study shows that LLMs can significantly reduce the workload of teachers in creating programming questions, but at the same time some limitations of the current use of large language models are also noted. Large language models and associated AI technologies are becoming indispensable tools in modern education, but their implementation requires strictly defined frameworks and human supervision to preserve academic integrity. 2. A. Flego, I. Mekterović (Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Retrieval-Augmented Generation System for Academic Knowledge Access  This paper presents the design, implementation, and evaluation of a Retrieval-Augmented Generation (RAG) system for academic knowledge access. The proposed system enables users to submit natural language questions and receive contextually relevant answers grounded in educational materials from multiple heterogeneous sources. The architecture combines semantic retrieval using a vector database with answer generation performed by large language models, supporting real-world student queries and the Croatian language. To ensure flexibility and cost-efficiency, the system supports both locally deployed open-source models and remote cloud-based models, while a comprehensive evaluation was conducted on a set of academic questions, assessing answer relevance, correctness, contextual grounding, and response latency. The results demonstrate that the RAG approach significantly improves answer quality in academic settings, while highlighting trade-offs between local and cloud-based models in terms of performance and accuracy. The findings confirm the suitability of retrieval-augmented techniques for educational applications and provide insights for future improvements, including broader knowledge coverage and user-driven content integration. 3. L. Brkić, H. Smontara, B. Franc, I. Mekterović (Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Application of Machine Learning and Deep Learning Methods for Solar Power Generation Forecasting   The paper discusses the application of machine learning and deep learning methods for short-term forecasting of solar power plant electricity production. The forecasts are based on historical production data and meteorological data available through external sources. Various machine learning models, including linear regression, support vector regression, random forests and XGBoost, as well as a deep learning model based on LSTM neural networks, are developed and evaluated. The forecast is performed for a time horizon of 72 hours with a resolution of 15 minutes. The model performance is assessed using standard error metrics, including MAE, RMSE, nMAE, nRMSE, sMAPE and the coefficient of determination R². The obtained results show that the machine learning and deep learning models achieve significantly better accuracy compared to the base physical model, especially in the short-term time horizon. The paper confirms the potential of applying these methods in predicting solar power plant production. 4. M. Kos, D. Pintar (Faculty of Electrical Engineering and Computing, Zagreb, Croatia), F. Vuić, K. Potočki (Faculty of Civil Engineering, Zagreb, Croatia) Comparative Assessment of Unsupervised Machine Learning Methods for Flood Wave Shape Analysis  Floods represent one of the most significant natural hazards, increasingly threatening both human safety and critical infrastructure, with their impacts being further amplified by climate change. In addition to flood magnitude, the shape of flood waves plays an important role in hydraulic design and flood risk assessment. This paper provides a comparison of popular unsupervised learning algorithms regarding their feasibility for flood wave shape analysis. The methods are applied on real-life historical data from hydrological stations in the Sava River Basin in Croatia. Three approaches are compared: K-means clustering, hierarchical cluster analysis (HCA), and self-organizing maps (SOM). K-means is used as a baseline method due to its widespread application in hydrological studies, while HCA and SOM are evaluated as alternative approaches capable of capturing structural similarities and nonlinear patterns in flood wave data. Flood events are represented using selected hydrological attributes and shape-related characteristics. Methods are compared and evaluated using clustering performance metrics, complemented by statistical analysis of flood wave characteristics through analysis of variance (ANOVA) as well as interpretation by domain experts. The results provide insight into the advantages and limitations of individual methods and support the selection of suitable data-driven approaches for flood wave classification. 5. M. Žagar, D. Delija, I. Branica, G. Sirovatka (Tehničko veleučilište Zagreb, Zagreb, Croatia) Cyber Forensics as a Constrained Form of Data Mining: A Methodological Perspective for Cyber Investigations  Cyber forensics is frequently presented as a deterministic technical discipline focused on reconstructing past events from digital evidence. However, legal admissibility requirements, most notably those articulated by the Daubert standard, demand explicit consideration of methodological validity, testability, and error rates. This paper argues that the analytical, inferential, and reporting phases of cyber forensic investigations are best understood as a constrained subset of data mining processes. By mapping core cyber forensic activities onto established data mining lifecycles, the paper demonstrates that cyber forensic reasoning closely aligns with data mining and data science workflows, particularly in its reliance on interpretation, correlation, and probabilistic inference. Recognizing and explicitly addressing uncertainty is therefore not a weakness, but a necessary condition for scientific rigor and legal credibility in cyber forensic practice. 6. M. Gulić, P. Turkalj (University of Rijeka, Faculty of Maritime Studies, Rijeka, Croatia), M. Valčić (University of Zadar, Maritime Department, Zadar, Croatia), N. Grubišić (University of Rijeka, Faculty of Maritime Studies, Rijeka, Croatia) Experimental Analysis of Pseudorandom Number Generators Considering Sequence Quality and Energy Consumption  Pseudorandom number generators are computer algorithms that produce sequences of numbers resembling random ones, making them essential for many applications. When selecting a suitable generator for a specific computer application, it is important to consider both the quality of the sequences and the energy efficiency of the algorithms. In this paper, various pseudorandom number generators are analyzed with emphasis on both the quality of the generated number sequences and their energy consumption. A software prototype implementing 27 generators was developed, and energy consumption was measured using the Wattmeter emos p5801 device. The quality of the generated numbers was assessed based on available literature and relevant research, while energy consumption was analyzed experimentally. The results show that some pseudorandom number generators achieve a high level of randomness, while others demonstrate better energy efficiency. Therefore, an additional ranking of the generators was conducted based on a weighted evaluation that considers both the quality of the generated sequences and energy consumption, aiming to provide a balanced assessment of the overall performance of the algorithms. It can be concluded that energy efficiency is a key criterion when selecting a number generator for a given application, along with the required quality of the generated sequences. 7. M. Tominac (Rimac Technology, Zagreb, Croatia), M. Krvavica, M. Vašak (Fakultet elektrotehnike i računarstva Sveučilišta u Zagrebu, Zagreb, Croatia), N. Čerkez (Rimac Technology, Zagreb, Croatia) Design of Experimental Procedures for Characterization of Battery Cells Aging  Aging of battery cells represents a major challenge in the development of reliable and long-lasting energy storage systems. Accurate characterization of battery cell aging is essential for understanding underlying degradation mechanisms, and prediction of the remaining useful life. To achieve a comprehensive characterization of aging phenomena, it is necessary to acquire experimental data that capture battery behavior across the full operational window, including near extreme state-of-charge (SOC) regions where the parameters of equivalent circuit models (ECMs) exhibit a steep rise. The experimental datasets are constructed using a combination of established industry-standard cycling protocols and newly designed profiles that deliberately extend operation at voltage limits to capture the electrochemical responses at SOC extremes. The contribution of this work is the systematic design of cycling protocols and in yielding high-resolution datasets with a comprehensive coverage across all SOC regions, including voltage limits. Well-constructed experimental datasets of this kind enable the development of accurate aging models, contribute to the comparability of research outcomes, and form the basis for reliable battery management and optimization of battery lifetime. 8. M. Nikolic, S. Stankovic (The Academy of Applied Technical and Preschool Studies, Nis, Serbia), M. Marjanovic (Singidunum University, Belgrade, Serbia) Integrating Generative AI into Business Intelligence for Sustainable Tourism Reporting  Sustainability has become a strategic priority in tourism, requiring analytical support for responsible decision-making and operational efficiency. However, many analytical systems in the tourism sector remain focused on visual summaries, offering limited support for narrative interpretation of performance indicators. Recent advances in generative artificial intelligence help business intelligence systems to move beyond dashboards toward automated analytical reporting. This paper proposes an integrated analytical framework that incorporates generative artificial intelligence into tourism sustainability analysis. Using public hotel booking data from Kaggle, operational indicators related to demand stability, occupancy utilization, seasonality, and cancellation behavior are derived through preprocessing, temporal aggregation, and normalization, establishing transparency and reproducibility. These indicators are then provided to an automated reporting component based on a large language model (LLM), which produces narratives aligned with the underlying data. Experimental results show high observed alignment between generated narratives and input indicators, with coverage exceeding 90%. The findings demonstrate that automation can be integrated into operational tourism analytics to enhance interpretability and support informed planning. 9. B. Agić, E. Mešković (University of Tuzla, Faculty of Electrical Engineering, Tuzla, Bosnia and Herzegovina) An Incremental Computing Approach for Geospatial Queries  Despite the long-standing demand for processing and utilizing geospatial data, existing approaches still show substantial limitations, including high computational latency, complex configuration requirements, or inadequate support for real-time analysis. Data streams containing a location (GPS coordinates) of a moving object at any time instant can be considered as frequent updates of moving objects’ positions which makes it possible to take advantage of incremental computing for their processing and management. Existing incremental computing engines have very limited support for geospatial data types and operations that would enable efficient processing of moving and stationary geospatial objects. This paper details the upgrade and usage of an incremental computing engine for geospatial data processing, and explores multiple connectivity mechanisms that facilitate the usage of geospatial data from both databases and data streams. With this approach it is possible to take the advantages of incremental computing queries for incremental calculation of current locations of moving objects. Proposed system prototype and implemented functionalities are illustrated through SQL-like queries.

Boris Vrdoljak (Croatia), Matteo Golfarelli (Italy), Mihaela Vranić (Croatia)

Marko Banek (Croatia), Matteo Golfarelli (Italy), Damir Pintar (Croatia), Mihaela Vranić (Croatia), Boris Vrdoljak (Croatia)

Mirta Baranović (Croatia), Ladjel Bellatreche (France), Ljiljana Brkić (Croatia), Alfredo Cuzzocrea (Italy), Matteo Francia (Italy), Enrico Gallinucci (Italy), Paolo Garza (Italy), Marko Gulić (Croatia), Luka Humski (Croatia), Igor Mekterović (Croatia), Danijel Mlinarić (Croatia), Sandro Skansi (Croatia), A Min Tjoa (Austria)
 

Prijava/Kotizacija:

Studentski popust se ne odnosi na studente doktorskog studija.

OBAVIJEST AUTORIMA: Uvjet za objavu rada je plaćanje najmanje jedne kotizacije po radu. Autorima 2 ili više radova, ukupna se kotizacija umanjuje za 10%.

Boris Vrdoljak
Fakultet elektrotehnike i računarstva
Unska 3
10000 Zagreb, Hrvatska

Tel.: +385 1 6129 532
Fax: +385 1 6129 915
E-mail: boris.vrdoljak@fer.hr 

Najbolji radovi bit će nagrađeni.
Prihvaćeni radovi bit će objavljeni u zborniku radova s ISSN brojem. Radovi na engleskom jeziku prezentirani na skupu bit će poslani za uključenje u digitalnu bazu IEEE Xplore.

Opatija je vodeće ljetovalište na istočnoj strani Jadrana i jedno od najpoznatijih na Mediteranu. Ovaj grad aristokratske arhitekture i stila već više od 180 godina privlači svjetski poznate umjetnike, političare, kraljeve, znanstvenike, sportaše, ali i poslovne ljude, bankare, menadžere i sve kojima Opatija nudi svoje brojne sadržaje. 

Opatija svojim gostima nudi brojne komforne hotele, odlične restorane, zabavne sadržaje, umjetničke festivale, vrhunske koncerte ozbiljne i zabavne glazbe, uređene plaže i brojne bazene i sve što je potrebno za ugodan boravak gostiju različitih afiniteta. 

U novije doba Opatija je jedan od najpoznatijih kongresnih gradova na Mediteranu, posebno prepoznatljiva po međunarodnim ICT skupovima MIPRO koji se u njoj održavaju od 1979. godine i koji redovito okupljaju preko tisuću sudionika iz četrdesetak zemalja. Ovi skupovi Opatiju promoviraju u nezaobilazan tehnološki, poslovni, obrazovni i znanstveni centar jugoistočne Europe i Europske unije općenito.

Detaljnije informacije se mogu potražiti na www.opatija.hr i www.visitopatija.com.