Search  English (United States) Hrvatski (Hrvatska)

innovative promotional partnershipArtificial Intelligence towards EU Multilingualism

Technical co-sponsorship

Hybrid Event
Event program
Thursday, 5/23/2024 9:00 AM - 12:45 PM,
Camelia 2, Grand hotel Adriatic, Opatija
9:00 AM - 12:45 PMDS - Znanost o podacima 
1.J. Manchev, M. Mirchev, I. Mishkovski (Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University in Skopje, Skopje, Macedonia)
Classification of Companies Using Graph Neural Networks 
Classification of companies into GICS categories can be addressed using Graph Neural Networks (GNN), by utilizing the different types of relationship between companies such as customer, supplier, partner, competitor, and investor. We use the Relato business graph data and compare the performances of several GNNs and a large language model like BERT that is trained only on the descriptions of the companies. Our goal is company classification into its corresponding category within the four tiers of the GICS hierarchy. Several architectures of GNNs are explored such as GCN, GraphSAGE and GAT, but also RGCN and RGAT that consider the edge type, or relationship between the companies. The main purpose is to reveal what kind of relationship between the companies is most valuable when determining the category of the company. The findings indicate that Graph Neural Networks (GNNs) enhance both classification performance and the understanding of collaboration patterns among companies, providing valuable insights for determining the industry in which these companies operate. This contrasts with the classification based solely on company descriptions using BERT.
2.N. Malivuk, V. Štrbac (University of Novi Sad, Faculty of Technical Science, Novi Sad, Serbia), S. Stefanović (University of Belgrade, Faculty of Philosophy, Belgrade, Serbia), L. Milić, M. Đoćoš (University of Novi Sad, Faculty of Technical Science, Novi Sad, Serbia), L. Adamov, B. Petrović (University of Novi Sad, Faculty of Medicine, Novi Sad, Serbia)
Quantitative Analysis of Ovicaprid Bone Tools from Early Neolithic Europe Using Dental CBCT Imaging 
Our approach involves conducting a quantitative analysis of Neolithic tools, namely weaning equipment, which have been previously documented and characterized by distinct human teeth marks. We utilized dental CBCT (cone beam computed tomography) imaging to scrutinize four samples in order to gain understanding of their internal volume and their significance in early feeding practices. Moreover, our intention was to provide a detailed examination of the composition of these bone tools. We employed a methodology that involved precise measurements of various aspects of the tools, including measuring the length of the tool, as well as the internal length, the maximal inner breadth, and the thickness of the recipient's rim. We utilized the onDemand3D software, which proved to be efficient in its ability to process DICOM files and employ image processing techniques to calculate the volume of the tools. Our investigation also uncovered details, such as damages to the rim and the presence of usage traces along the edge of the recipient. This study not only makes a significant contribution to the understanding of weaning practices during the Early Neolithic period, but it also serves as a prime example of the application of advanced imaging technologies in the field of archaeological research.
3.P. Zupančič (Faculty of Information Studies, Novo Mesto, Slovenia), P. Panov ( Jožef Stefan Institute, Ljubljana, Slovenia)
Anomaly Detection in Time-Series Employee Absence Data: A Case Study 
This paper provides an initial exploration of employee absence data for anomaly detection. Utilizing data collected from the MojeUre system, which aggregates employee data from diverse companies, our objective is to uncover hidden patterns and anomalies associated with absences. In this paper, we employ various anomaly detection techniques to identify and characterize unusual patterns in absence data. The comparative analysis in this paper offers valuable initial insights for organizations aiming to leverage data analytics for workforce management and strategic decision-making, particularly in the context of anomaly detection.
4.S. Paunkoska, G. Mirceva (Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University in Skopje, Skopje, Macedonia)
Hate Speech on Social Platforms through the Application of ML and NLP Methods 
Hateful behaviour on social platforms has recently become a topic of interest for many researchers. Users experience online encounters with instances of hate speech on a daily basis. This paper investigates how using modern machine learning and natural language processing techniques and methods make computer systems enhance their intelligence to effectively recognize words indicative of hate speech or insults. A performance comparison is conducted using an extensive dataset of publicly available posts, evaluating traditional classifiers against classifiers that rely on deep learning. The results indicate that the overall success of the model is not solely determined by the choice of classifier, but also by factors such as pre-processing of textual data and the accurate configuration of parameters.
5.E. Zolota, V. Hasić, A. Mević, A. Delić, S. Krivić (Faculty of Electrical Engineering, University of Sarajevo, Sarajevo, Bosnia and Herzegovina)
Predictive Analysis of Sarajevo's AQI using Machine Learning Models for Varied Granularity and Timeframes 
This study scrutinizes five years of Sarajevo's Air Quality Index (AQI) data using diverse machine learning models—ARIMA, SARIMA, Prophet FB, and LSTM—to forecast AQI levels. Focusing on various prediction frames (per day, hour, week, and month), we evaluate model performances and identify optimal strategies for different temporal granularities. Our research unveils subtle insights into each model's efficacy, shedding light on their strengths and limitations in predicting AQI across varied timeframes. This research presents a robust framework for automatic optimization of AQI predictions, emphasizing the influence of temporal granularity on prediction accuracy, automatically selecting the most efficient models and parameters. These insights hold significant implications for data-driven decision-making in urban air quality control, paving the way for proactive and targeted interventions to improve air quality in Sarajevo and similar urban environments.
6.I. Vasileska, P. Tomšič, L. Kos, L. Bogdanović (Faculty of Mechanical Engineering, University of Ljubljana, Ljubljana, Slovenia)
Unveiling Performance Insights and Portability Achievements between CUDA and SYCL for Particle-in Cell Codes on Different GPU Architectures 
The HPC systems worldwide are getting more powerful with the combination of CPU, GPU, and other accelerators (e.g., FPGAs and Quantum Processors). Many programming frameworks mainly offer excellent support portability to the existing scientific codes to use the exascale HPC systems. This study evaluates the performance and portability of CUDA and SYCL for one of the most used plasma kinetic simulation codes Particle-In Cell (PIC). The PIC codes are numerical modelling tools used for handling the extreme nonlinear methods in fusion devices. The experimental work showed that accelerating the PIC code with CUDA and SYCL achieve similar performance on NVIDIA devices, the latter demonstrated remarkable code portability to other GPU architectures. This brief study highlights the potential of SYCL as a viable solution for achieving both performance and portability in the heterogeneous computing ecosystem.
7.B. Kolarek (Institut Ruđer Bošković, Zagreb, Croatia), L. Gamulin (Hrvatski restauratorski zavod, Zagreb, Croatia), D. Davidović (Institut Ruđer Bošković, Zagreb, Croatia)
Cultural Heritage on HPC - Creating High Resolution 3D Models Using Photogrammetry 
In this presentation we will explore the application of High Performance Computing (HPC) in the creation of high resolution 3D models using photogrammetry for the preservation of cultural heritage. Photogrammetry is a non-invasive technique that uses multiple photographs enabling the creation of detailed 3D models that capture the intricate details of cultural heritage artefacts. This paper will introduce a workflow that utilises open source software on HPC systems to create high resolution 3D models from photogrammetry data. Using real-world examples of cultural heritage assets, we will demonstrate the benefits of HPC in creating high-resolution 3D representations that can be used for preservation, documentation and virtual exploration.
8.F. Fetaji, S. Gievska, K. Trivodaliev (Faculty of Computer Science and Engineering, Skopje, Skopje, Macedonia)
A Survey of Graph Neural Network Architectures in Ligand Binding Affinity Prediction Models 
Ligand affinity prediction plays a pivotal role in drug discovery, influencing the efficiency and success of drug development processes. Traditional methods struggle in accurately capturing the complex interactions within molecular structures, prompting the exploration of advanced techniques such as Graph Neural Networks (GNNs). This paper provides a comprehensive analysis of GNNs in the context of ligand affinity prediction, exploring their architecture, applications, and potential impact on revolutionizing drug discovery. Our findings reveal that GNNs offer significant improvements over traditional computational methods, particularly in handling the dynamic and complex nature of molecular interactions. We highlight innovative GNN architectures that have shown notable success in predicting ligand binding affinities, such as heterogeneous graph representation and attention mechanisms. The implications of these advancements suggest a paradigm shift in drug discovery, where GNNs can lead to more accurate predictions and accelerate the identification of potential drug candidates. This study underscores the transformative potential of GNNs in enhancing predictive accuracy and efficiency in drug development.
9.A. Krndžija, A. Kodžaga (International Burch University, Sarajevo, Bosnia and Herzegovina)
Real Estate Price Prediction 
The ability to predict real estate prices correctly has become essential for establishing the complexities of the constantly evolving real estate market. This research paper performs an in-depth evaluation of the most recent developments in this field, combining knowledge obtained through reading of a wide range of scientific publications. We analyze the complex study of several machine learning models, including a range of regression methods and time-series models, which are all used competitively to predict real estate prices. Our primary goal is to provide a nuanced understanding of the many ways used to approach this challenging issue by revealing the smallest details that are essential to real estate price prediction through this thorough review. Looking into the strengths and limits present in each model, our study not only gives an excluding review but also gives significant insights that add significantly to the extending discussion in real estate analytics. This study helps to improve the prediction models that are already in place and creates the foundation for strategic planning and well-informed decision-making in the dynamic real estate industry by pushing the limits of our understanding.
10.L. Prpić, P. Bago (Faculty of Humanities and Social Sciences, Zagreb, Croatia)
Exploring Linguistic Entropy: Correlating Entropy Values and Language Resources in the European Union 
The paper provides examination of linguistic entropy, a statistical metric that measures information content in textual data. Investigation extends beyond overall entropy assessments, delving into conditional entropies and considering the significance of word order. The central inquiries revolve around establishing relationships between the official languages of the European Union and their respective entropy values, with an intriguing exploration into entropy correlation based on linguistic characteristics of languages. The minimum corpus size required for this topic is identified. The amount of entropy depends on vocabulary size, therefore, for the consistency of the results, parliamentary discussions and debates are the topics of all corpora (source: Clarin and Sketch Engine). This approach, encompassing entropy analysis and resource development, combining linguistics, information theory and computer technology contributes to a comprehensive understanding of linguistic diversity within the European Union. Another aspect of this paper involves an exploration of the current state of language technology for the official EU languages, both low-entropy and high-entropy languages. The attempt will be made to establish a correlation between the quantity of produced language resources for a given language and its classification as either high-entropy or low-entropy.
Thursday, 5/23/2024 2:00 PM - 7:00 PM,
Camelia 2, Grand hotel Adriatic, Opatija
2:00 PM - 7:00 PMBE - Biomedicinsko inženjerstvo 
1.U. Marhl (Institute of Mathematics, Physics and Mechanics, Ljubljana, Slovenia), T. Sander (Physikalisch-Technische Bundesanstalt, Berlin, Germany), V. Jazbinšek (Institute of Mathematics, Physics and Mechanics, Ljubljana, Slovenia)
Using a Limited Number of Sensors in MEG or the Feasibility of Partial Head Coverage OPM MEG 
Since the fields produced by the brain are very weak, the neuroimaging technique magnetoencephalography (MEG) requires magnetometers with high sensitivity. Until recently, the only suitable sensor was the SQUID magnetometer, but in the last few years optically pumped magnetometers (OPMs) with comparable sensitivity have emerged. These sensors have the advantage, that they do not need cooling with a cryogenic liquid. As OPMs are still expensive, many laboratories perform OPM MEG with only partial coverage of the head. We, therefore, examined whether a limited number of magnetometers are sufficient for MEG measuring a focal activity such as the auditory evoked field. In more detail, we checked how the number of sensors affects the spatial resolution of the source localization method in MEG. To calculate the optimal sensor layout with a limited number of sensors we used the geometry of a whole-head SQUID MEG system. The optimal sensor layout was determined with the limited lead selection (LLS) algorithm. Additionally, we showed that this algorithm can be used to extend the measurement area for better visualization or to correct faulty channels if we have prior measurements with an extended set of sensors.
2.E. Jovicic, A. Jovic, M. Cifrek (University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia)
Impact of EEG Signal Preprocessing Methods on Machine Learning Models for Affective Disorders 
Affective disorders belong to a group of psychiatric disorders that are diagnosed according to the criteria of standardized diagnostic manuals. The diagnostic protocol consists of assessing the patient's symptoms, but to date, there are no methods to objectively evaluate or measure them. Electroencephalography (EEG) is a non-invasive brain electrical activity measuring technique. Current research mainly focuses on the use of EEG data and feature extraction, machine learning (ML), and deep learning (DL) to classify affective disorders. In this paper, the focus is on measuring the impact of preprocessing EEG signals on ML models for affective disorders. The impact of the following preprocessing methods are evaluated: signal filtering, independent component analysis (ICA), and canonical correlation analysis (CCA). The methods are assessed on a dataset consisting of EEG signals from 70 subjects diagnosed with affective disorders and 35 healthy subjects. After preprocessing, 570 features are extracted for each subject and several ML models are used for classification. CCA provided the best results compared to the other methods, with the highest F1 score of 0.9756. CCA should be considered as a beneficial preprocessing method to potentially improve classification results when building complex models for EEG data.
3.A. Šerifović Trbalić, A. Hasić, E. Skejić, N. Demirović (University of Tuzla, Tuzla, Bosnia and Herzegovina)
Seizure Detection based on EEG Signals and Deep Learning 
Epilepsy represents a neurological disorder of the brain characterized by repeated seizures. These are sudden abnormality in the brain's electrical activities that temporarily affect normal brain function. Electroencephalogram (EEG) is one of the main diagnostic tools for monitoring the brain activity of patients with epilepsy. Typically, the detection of epileptic activity is carried out by an expert by analyzing the EEG recordings, but this is a difficult, error prone and time-consuming task. In order to get timely and accurate automatic detection of seizure, various approaches based on both conventional and deep learning techniques were proposed in the literature. The aim of this paper is to present a framework for the automatic detection of epileptic seizure based on the functional connectivity matrix obtained from EEG signals and deep learning. A convolutional neural networks (CNN) were employed because of their capability to learn patterns of neural activities based on brain connectivity represented by connectivity matrix. Obtained results are very promising indicating a potential of this approach as an efficient tool for automated seizure detection based on EEG data.
4.L. Jelić, M. Cifrek, V. Lešić (University of Zagreb Faculty of Electrical Engineering and Computing, Zagreb, Croatia), P. Orepić (University of Geneva, Neural Dynamics Lab, Department of Basic Neurosciences, Geneva, Switzerland)
Enhancing the Self-Other Voice Discrimination Procedure for Clinical Applications 
Self-voice plays an important role in our everyday lives. Apart from being used for communication, our own voice defines our identity, as it is the sound that is most intimately connected to ourselves. Disturbances in self-voice recognition have been related to certain psychotic symptoms such as auditory verbal hallucinations, colloquially called ‘hearing voices’, and to a distorted sense of self more generally. Recent research has indicated that a specific self-other voice discrimination procedure combining psychophysics, voice morphing technology, and air- and bone- conducted voice stimuli can be of clinical significance as a biomarker for detecting pathological alterations in self consciousness. However, there are several limitations with the existing approach, mainly in not being simple to use and difficult to apply in different clinical contexts. To address these drawbacks, we adapted the current methodology and developed a self-other voice discrimination solution with a graphical user interface, automated voice morphing with personalized voice selection tool, and automated results generation. This improves the usability of the task, shortens the procedure duration, and provides a patient-specific approach. This paper demonstrates the technological advantages and scientific potential of the new methodology, with sample data from two different participant groups in clinical and non-clinical settings.
5.M. Memon, J. Jadebeck, M. Osthege (Forschungszentrum Juelich GmbH, Juelich, Germany), A. Wendler (German Aerospace Center, Cologne, Germany), D. Kerkmann (Helmholtz Centre for Infection Research, Braunschweig, Germany), H. Zunker (German Aerospace Center, Cologne, Germany), W. Wiechert, K. Nöh, J. Goebbert, B. Hagemeier (Forschungszentrum Juelich GmbH, Juelich, Germany), M. Riedel (University of Iceland, Reykjavik, Iceland), M. Kühn (German Aerospace Center, Cologne, Germany)
Automated Processing of Pipelines Managing Now- and Forecasting of Infectious Diseases 
When faced with the challenge of now- and forecasting infectious diseases, multiple data sources and state-of-the-art models have to be considered. Automatic aggregation, processing, and publishing to relevant data sinks is paramount to achieving consistent, reproducible, and timely results given daily-reported data. To facilitate scientific collaboration and reproducibility of workflows, open and extensible architectures for compute pipelines are required. In this research, we devise an architecture realizing the seamless management and processing of reproducible pipelines. Our case-study is a daily pipeline for nowcasting the state of SARS-CoV-2 in Germany based on public data and state-of-the-art models implemented in the simulation software MEmilio. The results of our pipeline are pushed to ESID (Epidemiological Scenarios for Infectious Diseases), a user interface to epidemiological simulations. To realize the given pipeline, a workflow management system is required to ensure pipeline processing and secure access to multiple heterogeneous data storages. For this purpose, we based our work on an open-source workflow management system- Apache Airflow, which provides the orchestration, coordination and management of complex connected tasks. S3 is utilized as an intermediate data storage service for sharing data between workflow steps and persisting experiment output. We provide a comprehensive view on our work on automated, end-to-end and reproducible pipelines, with detailed commentary on use case, and its realization.
6.D. Katović (University of Zagreb Faculty of Kinesiology, Zagreb, Croatia), T. Bronzin (CITUS, Zagreb, Croatia), M. Horvat (University of Zagreb Faculty of Electrical Engineering and Computing, Department of Applied Computin, Zagreb, Croatia), B. Prole, A. Stipić (CITUS, Zagreb, Croatia), N. Jelača, I. Pavlović, K. Pap (University of Zagreb Faculty of Graphic Arts, Zagreb, Croatia)
The Use of AI in Human Pose Estimation Applications in Kinesiology: Taxonomy of Algorithms, Models, and Evaluation Methods 
Kinesiology and its related disciplines (kinanthropometry, biomechanics, kinesiological rehabilitation, sports games) symbiotically attract new computer technologies, implementing them in the form of effective tools for analysis, diagnosis, and assessment of the state of the subject or team. The use of neural networks and computer vision integrated into pose estimation technology can be used by kinesiology as an applicable integration of knowledge from the domain of movement analysis, analysis of sports games, rehabilitation, and education into an environment of computer-transformed visual information and patterns into a form suitable for further analytical processes. The paper represents a structured review of algorithms, models, and evaluation methods for recognizing and monitoring human movement, single or multi-person, in real-time using human pose estimation.
7.I. Putnik, N. Petrović (University of Novi Sad, Faculty of Techical Sciences, Novi Sad, Serbia), K. Joseph, Kuala Lumpur, Malaysia), A. Thiha (University of Malaya, Faculty of Engineering, Centre for Innovation in Medical Engineering(CIME), Kuala Lumpur, Malaysia), M. Vejin, S. Kojić, G. Stojanović (University of Novi Sad, Faculty of Techical Sciences, Novi Sad, Serbia)
Microfluidic CD Fabrication for Electrochemical Analysis of pH-varied Sweat 
This study presents a systematic redesign of a microfluidic compact disc (CD) that showcases six distinct chamber subsystems. This type of a subsystem is specifically designed to facilitate concurrent sample loading, mixing, and analysis. Departing from the original focus on saliva as a diagnostic medium, this study shifts towards the utilization of sweat. Modifications are made to the electrode designs to ensure accurate comparison and electrochemical analysis. The primary aim of this investigation is to explore the variations in electrochemical properties that occur within dilution of sweat samples as well as varying pH levels. This is important as specimens can vary greatly in their composition and complexity. By contrasting the electrochemical impedance characteristics of sweat samples with different pH levels, this study aims to shed light on potential foundations for the early diagnosis and monitoring of skin diseases. Preliminary findings from this extension of the study reveal distinct conductance patterns. These findings provide valuable insights and further support the notion that the redesigned microfluidic CD platform possesses the capability for multiplex processes. This, in turn, paves the way for future research in the field of theranostics and the electrochemical analysis of diverse sweat compositions at the point of care.
8.I. Vujaklija, S. Biđin, M. Šikić (Faculty of Electrical Engineering and Computing, Zagreb, Croatia)
Comparative Analyses of Nanopore-based Unsupervised Learning Methods in Epitranscriptomics 
RNA modifications, collectively known as epitranscriptome, are present in all kingdoms of life. To date, more than 300 types of RNA modifications have been reported. Detecting a wide range of epitranscriptomic modifications simultaneously, remains an elusive goal for supervised learning methods, and unsupervised approaches are needed. Here, we compared two main types of unsupervised approaches based on the third generation nanopore sequencing: signal-based and basecaller-errorbased. Their performance was evaluated on the newly developed benchmark test dataset, constructed from recently released data comprising 25S ribosomal subunit of S. cerevisiae. The advantage of the rRNA test dataset used in our study is that it consists of native molecules containing many different modification types.
9.I. Tomasic (Mälardalen University, Västerås, Sweden), A. Prkic (University Hospital of Split, Split, Croatia), A. Lesin, D. Kalibovic Govorko, I. Medvedec Mikic ( School of Medicine, University of Split, Split, Croatia)
Respiratory Rate Estimation from a Single Lead ECG Obtained During Dental Surgery 
Respiratory rates were estimated from single lead ECG measurements, obtained on 30 patients during dental surgery, by using the Toolbox of Respiratory Rate Algorithms, from the Respiratory Rate Estimation (RRest) project. The purpose was to find the best performing among more than 300 algorithms implemented in the toolbox. Since the dataset does not contain a reference respiratory signal, respiratory rate extraction algorithms were compared with respect to the mean respiratory rate calculated from applying all the algorithms. The investigation found one algorithm that estimates respiratory rates within tolerance of ±𝟐 breaths/min in at least 50% of the recordings’ lengths, for 27 of the total 30 measurements. The selected algorithm has not satisfied the criteria on 3 patients, possibly because of the big variation in the shape of the recorded ECGs, but also because the patch ECG device was not in the position where ECG amplitude modulation, caused by respiration, is expected. The selected algorithm is the smart fusion of respiratory rates obtained by peak detection and zero-crossings of respiratory signals, that are calculated from ECG features.
10.I. Kuzmanov, E. Zdravevski, P. Lameski, B. Stojkoska, A. Madevska Bogdanova (Ss. Cyril and Methodius University – Skopje, Faculty of Computer Science and Engineeing - FCSE, Skopje, Macedonia)
A Study on Appropriate Segment Length for Generalized Cuff-less Blood Pressure Estimation from ECG Features 
Blood pressure (BP) refers to the pressure exerted on the blood vessels as blood travels through the body. Our ultimate goal is to build a stable model for BP estimation as part of a triage process. In this study, we experiment to determine a suitable signal segment only from ECG signals, to ensure a fast and reliable process of the BP estimation. The used dataset contains only high-quality ECG and ABP signals extracted from MIMIC II and MIMIC III databases by our methodology. It was processed three times using similar ML methodologies, with different segment lengths. Three different datasets are generated using a non-overlapping window with a size of 8, 15, and 30 seconds, with the same ECG features. Several linear and nonlinear Machine Learning models are built on these datasets, and their results are compared. Our best results were obtained by a LightGBM regression model trained on the 30-second dataset. The model achieves MAE of 10.87, 6.55, and 7.29, and RMSE of 14.49, 8.68, and 9.68 for SBP, DBP, and MAP, respectively. The results of our experiment indicate that a duration of 30 seconds is the minimum length that provides informative features, fulfilling the need for real-time delivery.
11.A. Angjelevska (Innovation Dooel, Skopje, Macedonia), M. Gusev (Ss. Cyril and Methodius University/Faculty of Computer Science and Engineering, Skopje, Macedonia), S. Gjorgjieva (Innovation Dooel, Skopje, Macedonia)
Classification of Hemoglobin A1c from Long and Extra-long Term Heart Rate Variability 
This study classifies Hemoglobin A1c (HbA1c) concentration from long- and extra-long-term heart rate variability (HRV) measurements and various machine learn ing (ML) models utilizing different datasets. Key metrics include SDNN, RMSSD(NN), NN50, and PNN50 under detailed window-oriented calculations, employing Average, Standard Deviation, and Concatenated methods for feature extraction. A comprehensive pre-processing phase within the ML pipeline ensures analytical robustness. The study systematically conducts patient-wise data splits and evalu ates classification performance across various ML models, contributing to a thorough analysis. Evaluation metrics such as sensitivity, specificity, precision, and different F1 scores guide this research in advancing the understanding of HbA1c regulation. The study aspires to establish optimal ML model training and evaluation configurations, contributing to the broader discourse on HbA1c classification. The best performing model reaches an F1 Score of 93.20%, and F1M of 92.76%, demonstrating its robustness and effectiveness over baseline models.
12.G. Temelkov (Innovation Dooel, Skopje, Macedonia), M. Gusev (Univ. Sts Cyril and Methodius, Skopje, Macedonia)
Leveraging Dataframe-Based Operations for Calculation of Heart Rate Variability 
This paper introduces a novel Heart Rate Vari ability (HRV) calculation strategy, diverging from traditional divide-and-conquer methodologies to dataframes utilizing the Polars library as an advanced data manipulation tool. We demonstrate significant improvements in computational performances, where the dataframe approach outperforms the iterative approach with speedup factors beyond 98 times for short-term HRV calculations and substantial reductions in processing times across various test cases. These en hancements underscore the potential of tailored dataframe manipulations in enhancing performance and adapting to complex data analysis challenges in HRV assessments.
13.N. Petrovski, M. Gusev, A. Kulakov (Ss Cyril and Methodius University, Faculty of Computer Sciences and Engineering, Skopje, Macedonia)
Atrial Fibrillation Detection using the Stars 2D Convolutional Neural Network 
Atrial Fibrillation is one of the riskiest potentials of heart failure in the recent escalating prevalence of cardiovascular diseases. Detecting such an irregularly irregular heart rhythm is a paramount challenge in modern Biomedical Computation. This research contributes to its resolution by introducing an advanced 2D Convolutional Neural Network model, trained on the Poincare plot of the differences between consecutive Beats Per Minute (BPM) values. Furthermore, we use the Gaussian Blur technique to enhance the model's capacity to generalize and augment its accuracy by forming a star-like plot similar to the famous Van Gogh's Starry Nights picture. Applying the standard Machine Learning window label comparison method our model achieves an impressive F1 Score of 95.55\%. Evaluating the duration-based assessment method the model achieves a remarkable F1 Score of 94.59\%.

Basic information:

Karolj Skala (Croatia), Aleksandra Rashkovska Koceva (Slovenia), Davor Davidović (Croatia)

Steering Committee:

Sven Lončarić (Croatia), Morris Riedel (Germany), Uroš Stanič (Slovenia), Matjaž Veselko (Slovenia), Yingwei Wang (Canada), Martin Žagar (Croatia)

Program Committee:

Enis Afgan (Croatia), Viktor Avbelj (Slovenia), Davor Davidović (Croatia), Matjaž Depolli (Slovenia), Simeon Grazio (Croatia), Marjan Gusev (North Macedonia), Vojko Jazbinšek (Slovenia), Jurij Matija Kališnik (Germany), Zalika Klemenc-Ketiš (Slovenia), Dragi Kocev (Slovenia), Gregor Kosec (Slovenia), Miklos Kozlovszky (Hungary), Lene Krøl Andersen (Denmark), Željka Mihajlović (Croatia), Panče Panov (Slovenia), Tonka Poplas Susič (Slovenia), Aleksandra Rashkovska Koceva (Slovenia), Karolj Skala (Croatia), Viktor Švigelj (Slovenia), Ivan Tomašić (Sweden), Roman Trobec (Slovenia), Roman Wyrzykowski (Poland)

Registration / Fees:

Price in EUR
Up to 6 May 2024
From 7 May 2024
Members of MIPRO and IEEE 243 270
Students (undergraduate and graduate), primary and secondary school teachers 130 150
Others 270 300

The discount doesn't apply to PhD students.

NOTE FOR AUTHORS: In order to have your paper published, it is required that you pay at least one registration fee for each paper. Authors of 2 or more papers are entitled to a 10% discount.


Karolj Skala
Rudjer Boskovic Institute
Center for Informatics and Computing
Bijenicka 54
HR-10000 Zagreb, Croatia



All submitted papers will pass through a plagiat control and blind peer review process with at least 2 international reviewers.

On the basis of reviewers' opinion and voting result from the conference attendance we will qualify the Best paper for the prize that will be awarded as a part of the final event at the DS-BE conference.

Accepted papers will be published in the ISSN registered conference proceedings. Papers presented at the conference will be submitted for inclusion in the IEEE Xplore Digital Library.
There is a possibility that the selected scientific papers with some further modification and refinement are being published in the following journals: Journal of Computing and Information Technology (CIT)MDPI Applied ScienceMDPI Information JournalFrontiers and EAI Endorsed Transaction on Scalable Information Systems.



Opatija is the leading seaside resort of the Eastern Adriatic and one of the most famous tourist destinations on the Mediterranean. With its aristocratic architecture and style, Opatija has been attracting artists, kings, politicians, scientists, sportsmen, as well as business people, bankers and managers for more than 170 years.

The tourist offer in Opatija includes a vast number of hotels, excellent restaurants, entertainment venues, art festivals, superb modern and classical music concerts, beaches and swimming pools – this city satisfies all wishes and demands.

Opatija, the Queen of the Adriatic, is also one of the most prominent congress cities in the Mediterranean, particularly important for its ICT conventions, one of which is MIPRO, which has been held in Opatija since 1979, and attracts more than a thousand participants from over forty countries. These conventions promote Opatija as one of the most desirable technological, business, educational and scientific centers in South-eastern Europe and the European Union in general.

For more details, please visit and

News about event
Currently there are no news
Patrons - random
ENT ZagrebA1 HrvatskaEkonomski fakultet RijekaTehničko veleučilište u ZagrebuHATZ