Irfan Fetahović

Journal of Scientific & Industrial Research (JSIR)

Feature Selection for Biomedical Data Classification: Statistical vs. Swarm Intelligence Methods

U. Marovac, A. Avdic, I. Fetahovic, L. Memic, N. Djordjevic, Z. Dolicanin, G. Babic

https://or.niscpr.res.in/index.php/JSIR/article/view/13842

Applying machine learning methods to large datasets with numerous features presents challenges in terms of training time and model complexity. Feature selection is crucial for reducing data dimensions, improving classification accuracy, and optimizing model interpretability. This study aims to enhance the classification of integrated biomedical data to identify thrombophilia diagnosis. The dataset consists of 71 features from 35 women (22 healthy, 13 with thrombophilia), and three classification algorithms (K Nearest Neighbors, Random Forest, Support Vector Machine) are used to evaluate model performance. Identifying key features related to thrombophilia diagnosis is performed using both filter methods and wrapper methods based on swarm intelligence algorithms. Those methods are analyzed and compared as potential approaches for the feature selection process. The wrapper method outperformed the filter methods for clinical and biological data, achieving a classification accuracy of 0.97 compared to 0.91, while selecting only 4 key features compared to 10. For demographic data, both methods produced the same classification accuracy (0.83), but the wrapper method reduced the number of features. These findings demonstrate that wrapper methods based on swarm intelligence algorithms improve model performance and facilitate more efficient data management, which holds significant practical applications for thrombophilia diagnostics. Additionally, the study highlights the advantage of applying the Bat Algorithm in the feature selection process for thrombophilia prediction, contributing to both the novelty and utility of the approach.

Biomedical data classification, Feature selection, Machine learning, Swarm intelligence

M22

Artificial Intelligence Conference Book of Abstracts

Swarm intelligence methods in feature selection for biomedical data classification

I. Fetahović, A. Avdić, U. Marovac

https://www.mi.sanu.ac.rs/~ai_conf/2025/AI_Conference_Book_of_Abstracts.pdf

Applying machine learning methods to large datasets with numerous features presents challenges related to training time, model complexity, and interpretability. Feature selection addresses these issues by reducing data dimensionality, improving classification accuracy, and enhancing interpretability. This study aims to improve the classification of integrated biomedical data for thrombophilia diagnosis. The dataset comprises 71 features collected from 35 women, including 22 healthy individuals and 13 diagnosed with thrombophilia. To evaluate model performance, three classification algorithms were applied: K-Nearest Neighbors (KNN), Random Forest (RF), and Support Vector Machine (SVM). Subjects were recruited from the Gynecology-Obstetrics Clinic of the University Clinical Center Kragujevac, encompassing both healthy pregnant women and patients with thrombophilia confirmed by molecular diagnostics. Key features related to thrombophilia diagnosis were identified using wrapper methods based on swarm intelligence algorithms. The results indicate that swarm intelligence-based wrapper methods substantially improve model performance and contribute to more efficient data management, which has important practical implications for thrombophilia diagnostics.

feature selection, machine learning, biomedical data classification, swarm intelligence

M34

Ethics in Online AI-based Systems

Data-driven educational decision-making model for curriculum optimization

E. Mekić, I. Fetahović, K. Kuk, B. Popović, and P. Čisar

https://www.sciencedirect.com/science/chapter/edited-volume/abs/pii/B9780443188510000020

This chapter explores the implementation of machine learning, artificial intelligence (AI), and big-data analyses in the decision-making model of curriculum optimization. Preparing the curriculum is only a small part of the overall process of planning and implementing the education cycle. Data obtained and analyzed with machine learning techniques from archives can provide insight into the past efficiency of teaching, help us to predict future trends, and change the curriculum according to the expected outcomes. However, implementation of these tools is limited due to the number of ethical challenges, such as security, possible stereotype-based decisions, and generalizations. In this chapter, we should explain the cycles of education planning and implementation, as well as the appropriate machine learning techniques used to support decisions and the ethical consequences of doing so.

Machine-learning, Artificial intelligence, Big data, Curriculum, Ethical optimization

M14

International Journal of Engineering Education

Innovative Approach to Teaching Distributed Systems in Education 4.0

S. Purkovic, I. Fetahovic, E. Mekic, G. Katipoglu, and S. Utku

https://www.ijee.ie/contents/c400524.html

This paper presents an innovative pedagogical approach for teaching university-level computer science courses, specifically distributed systems. Integrating concepts from computer science and educational sciences, the approach combines large language models (LLMs), DevOps tools, and the educational framework based on agile approach in which students take responsibility for organizing their own learning process (eduScrum) within the vision of aligning education with the demands of the 21st century (Education 4.0). Over four academic years, the new approach was compared to traditional teaching methods. The course was structured into three sprints, each encompassing theoretical and practical tasks. ChatGPT was utilized in example solutions for code generation and debugging, while Git repositories supported practical programming exercises. Efficiency was measured through qualitative and quantitative analysis, indicating a marked improvement with the new methodology. The study demonstrates the potential of modern technologies to create dynamic, effective learning environments and suggests a pathway for updating computer science and software engineering education to keep pace with technological advancements.

eduScrum; distributed systems; large language models; Education 4.0; DevOps

M22

8th International Conference CONTEMPORARY PROBLEMS IN MATHEMATICS, MECHANICS AND INFORMATICS CPMMI Book of Abstracts

Ethical Guidelines for Open Learning Analytics

I. Fetahovic, E. Mekic, K. Kuk, S. Purkovic, E. Dolicanin

http://www.dunp.np.ac.rs/wp-content/uploads/2022/cpmmi2024/Book%20of%20Abstracts%2C%20CPMMI%202024.pdf

Learning analytics (LA) is an interdisciplinary field between data science, artificial intelligence (AI), statistics, learning sciences, and psychology. Open learning analytics (OLA) is a research area focused on integrating both educational data and heterogeneous LA approaches on an open platform. In this research the most important ethical issues in OLA, such as data privacy, transparency, bias, and fairness, are briefly discussed. Ethical guidelines for OLA are proposed and divided into three parts, for three distinct stakeholders: developers, deployers, and end users.

learning analytics, ethics, artificial intelligence

M34

8th International Conference CONTEMPORARY PROBLEMS IN MATHEMATICS, MECHANICS AND INFORMATICS CPMMI Book of Abstracts

Comparison of Feature Selection Methods for Biomedical Data Classification

U. Marovac, A. Avdic, I. Fetahovic, L. Memic, N. Djordjevic, Z. Dolicanin

http://www.dunp.np.ac.rs/wp-content/uploads/2022/cpmmi2024/Book%20of%20Abstracts%2C%20CPMMI%202024.pdf

Prior to applying machine learning techniques to huge datasets with a large number of features, it is crucial to identify and select the most significant ones in order to reduce the size of the dataset and improve model performance. In this research, both filter and wrapper-based metaheuristic approaches were used for biomedical data classification. The obtained results demonstrate that applying feature selection improves the model’s performance, or at least provides the same results while reducing the size of the data set and making data collection easier.

feature selection, biomedical data classification, machine learning

M34

SCIENTIFIC PUBLICATIONS OF THE STATE UNIVERSITY OF NOVI PAZAR SER. A: APPL. MATH. INFORM. AND MECH

Ethical Challenges in Open Learning Analytics

I. Fetahovic, E. Mekic, K. Kuk, B. Popovic, E. Dolicanin

https://scipub-a.np.ac.rs/archive/2023/vol-15-no-2-2023/ethical-challenges-in-open-learning-analytics.html

Open learning analytics (OLA) is a new research field focusing to create an open platform for integration of heterogeneous learning environments. The central notion in OLA is openness, and it relates to architectures, processes, access, and datasets. Learning analytics usually employs artificial intelligence and machine learning algorithms to provide information and predictions, thus enhancing learning experience. AI-based systems provide many benefits, but their implementation raises ethical concerns due to a wide range of possible negative consequences. In this paper a comprehensive list of ethical issues in OLA is discussed along with mitigation procedures.

open learning analytics, ethics, artificial intelligence, machine learning, education

M52

The 9th Conference with International Participation on Knowledge, Menagement and Informatics, Kopaonik, Serbia, Book of Proceedings

An Implementation of Microservices Architecture Using Heterogenous Tools and Libraries

I. Fetahović, A. Milanović, A. Avdić, and L. Memić

https://kmi.vtsns.edu.rs/KMI_2023/radovi/1-KMI_Informatika/10.pdf

Microservices architecture is a new paradigm for creating modern software systems. It structures an application as a collection of loosely coupled, fine-grained services which communicate through lightweight protocols. This approach facilitates development parallelization by allowing small, autonomous teams or individuals to independently develop, deploy and expand services. In this paper, microservices architecture is discussed and the Netflix Clone application is presented, developed using this architectural style. The application demonstrates many benefits of microservices, such as increased modularity, easier understanding, development and testing, as well as flexibility through the use of different programming languages, tools and libraries.

microservices, software architecture, fine-grained services

M33

International Review

Kirkpatrick’s Model of Evaluation applied under Covid 19 conditions

S. Zejnelagic, I. Fetahovic, S. Moretic-Micic

https://scindeks.ceon.rs/Article.aspx?artid=2217-97392203030Z

Evaluation is of high importance in any format present in the educational process at any level. In the paper, Kirkpatrick’s model of evaluation is discussed under Covid-19 conditions. The evaluation process under these circumstances and the satisfaction level of both students and tutors are described, as well as changes that occurred during the pandemic. The focus is on virtual classroom activities and the importance of evaluation under Covid-19 conditions, including time experience of both students and tutors in this virtual learning environment.

e-learning; Covid-19; a virtual language classroom; learning satisfaction

M23

SCIENTIFIC PUBLICATIONS OF THE STATE UNIVERSITY OF NOVI PAZAR, SER. A: APPL. MATH. INFORM. AND MECH

The Process of Making, Organizing and Using Thematic Layers in GIS

D. I. Radovic, I. S. Fetahovic, E. C. Dolicanin

https://scipub-a.np.ac.rs/archive/2022/vol-14-no-2-2022/the-process-of-making-organizing-and-using-thematic-layers-in-gis.html

As Geographic Information System (GIS), as part of the Geoinformation Sciences, deals with a complete system of creating, handling, storing, manipulating, using analysis and presenting information related to the Earth’s surface, it is in the direct function of solving fundamental issues arising from that process. The organization of data is according to the system of thematic data layers. The paper presents specific indicators created in the process of GIS development, including qualitative, quantitative and visual representation of the real environment, as well as the results of multicriteria thematic and topological analyses.

GIS, thematic data layers, data types, thematic and topological analyzes, visualization

M52

7th International Conference CONTEMPORARY PROBLEMS IN MATHEMATICS, MECHANICS AND INFORMATICS CPMMI Book of Abstracts

Hyperparameter tuning in machine learning algorithms

I. Fetahović, E. Dolićanin, D. Radović, and L. Memić

—

Most machine learning algorithms have a certain number of parameters which cannot be learned. These parameters are called hyperparameters and they must be carefully chosen because they significantly impact overall performance of the algorithm. In this paper different approaches in hyperparameter optimization and tuning are analyzed, including popular libraries and frameworks which implement hyperparameter optimization algorithms, as well as methods and techniques in deep learning.

—

M34

7th International Conference CONTEMPORARY PROBLEMS IN MATHEMATICS, MECHANICS AND INFORMATICS CPMMI Book of Abstracts

The Process of Making, Organizing and Using Thematic Layers in GIS

D. Radović, I. Fetahović, and E. Dolićanin

—

As Geographic Information System (GIS) deals with a complete system of creating, handling, storing, manipulating, using analysis and presenting information related to the Earth’s surface, it is in the direct function of solving fundamental issues arising from that process. The research includes the formation of GIS for different types of research for the Tara Mountain area and the Skadar Lake region. Various thematic data layers were created, and the results of multicriteria thematic and topological analyses are presented.

—

M34

Zbornik radova LXVI Konferencija za elektroniku, telekomunikacije, računarstvo, automatiku i nuklearnu tehniku, ETRAN

Analiza performansi kooperativnog diverziti sistema u kompozitnom fedingu modelovanom odnosom α-μ i gama raspodela

E. Mekić, I. Fetahović, i Edin Dolićanin

—

U ovom radu je izvedeno novo opšte, jednostavno rešenje u zatvorenom obliku za funkciju gustine verovatnoće odnosa proizvoda slučajnih promenljivih predstavljenih α-μ i Gama raspodelama i slučajne promenjive predstavljene Gama raspodelom. Ova rešenja se primenjuju u analizi performansi komunikacionih sistema sa kooperativnim diverziti sistemom koji se koristi za poboljšanje prijema signala na čije anvelope utiče brzi i spori feding, dok na anvelopu kokanalne interference utiče samo brzi feding.

α-μ raspodela; Gama raspodela; feding; kooperativni diverziti

M63