mgr inż. Aleksander Wojtowicz

ORCID: 0009-0006-4514-4608

The topic of my dissertation is: “Interval-valued fuzzy methods and aggregation functions for feature selection in information systems”.

I am a doctoral student in information and communication technology (informatyka techniczna i telekomunikacja) at the Doctoral School of the University of Rzeszów and a research and teaching assistant in the Department of Artificial Intelligence at its Institute of Computer Science. My supervisor is dr hab. inż. Wiesław Paja, prof. UR. I work on feature selection, classification and the use of fuzzy methods to describe uncertainty in data.

In some datasets, there are many more variables than observations. A gene expression dataset, for instance, may contain thousands of measurements per sample. A model can appear to perform well on such data even when the variables it relies on change substantially from one sample to another. My dissertation asks how to select a smaller set of variables and how to tell whether the resulting ranking is dependable.

I use intervals to represent how the estimated importance of a feature changes across subsets of the data. An interval captures the range of scores a feature receives. Aggregation functions then combine the available scores so that the ranking can reflect both the level of importance and how consistently it is estimated. I study how the choice of importance measure and aggregation function affects which features end up at the top of the ranking.

Three studies examine this approach directly. In the IEEE FUZZ-IEEE 2025 paper, my co-authors and I proposed the Interval-Valued Weighted Feature Ranking (IVWFR) method, which combines feature importance measures obtained across data folds into interval-valued scores. In Applied Sciences, we extended that method with weighted interval aggregation and compared it with nine established approaches on ten synthetic and five real datasets, using statistical tests to assess the significance of the differences. My work with Wiesław Paja on the Adaptive Interval Feature Selector (AIFS), published in the ISD 2026 proceedings, addresses how to carry out feature selection with fewer repeated model fits: the method is non-iterative, isolates selection from the validation partitions to avoid data leakage, and was evaluated on 25 datasets, reaching an average balanced accuracy of 0.8334 and an average Kuncheva stability index of 0.6904, while remaining usable when 30% of the data are missing.

I compare methods on three questions. How much does the selected set change when the data change? How well does a classifier perform using that set? How much computation is required? My earlier work on ensemble classifiers for microarray data provides a starting point for these comparisons. I also work with FTIR and Raman spectroscopy, where identifying which parts of a spectrum influence a prediction can help researchers interpret a classifier.

Alongside my dissertation, I contribute to the NCN OPUS 27 project “Identification and measurement of systemic threats to the insurance sector arising from climate risk in the light of sustainable development policy: new tools based on machine learning methods” (registration no. 2024/53/B/HS4/03258), carried out at the Cracow University of Economics under the direction of dr hab. Anna Denkowska, prof. UEK. Our ISD 2026 paper examined how macroeconomic shocks — inflation, interest rates and GDP — relate to technical provisions and liquidity in the insurance sectors of Poland, Czechia and Hungary in 2016–2024, comparing Ridge regression, Random Forest, Gradient Boosting and panel fixed-effects models. This work uses machine learning to investigate a different kind of question, where the variables of interest describe the economy and the financial position of insurers.

Keywords: feature selection, feature importance, feature ranking stability, interval-valued fuzzy sets, aggregation functions, machine learning, high-dimensional data, interpretable models.

Publications:

Wojtowicz, A., & Paja, W. (2026). Adaptive Interval Feature Selector Based on Aggregation Functions for Resilient and Computationally Optimal Learning. In M. Valenta, B. Mannová, R. Pergl, A. Przybylek, M. Lang, H. Linger, C. Schneider, N. Iivari, & E. Insfran (Eds.), Making ISD Sustainable: Reloaded with AI and Automation (ISD2026 Proceedings) — 34th International Conference on Information Systems Development, Prague, Czechia, 2–4 September 2026. Full Paper. Prague: Czech Technical University in Prague. ISBN 978-80-01-07585-2. https://doi.org/10.62036/ISD.2026.40

Wojtowicz, A., Rak, E., Denkowska, A., & Wanat, S. (2026). Machine Learning-Driven Assessment of Systemic Risk and Macroeconomic Shocks in CEE Insurance. In M. Valenta et al. (Eds.), Making ISD Sustainable: Reloaded with AI and Automation (ISD2026 Proceedings) — 34th International Conference on Information Systems Development, Prague, Czechia, 2–4 September 2026. Short Paper. Prague: Czech Technical University in Prague. ISBN 978-80-01-07585-2. https://doi.org/10.62036/ISD.2026.77

Wojtowicz, A., Paja, W., & Bentkowska, U. (2025). Feature Importance Ranking Using Interval-Valued Methods and Aggregation Functions for Machine Learning Applications. Applied Sciences, 15(22), 12130. https://doi.org/10.3390/app152212130

Wojtowicz, A., Bentkowska, U., & Paja, W. (2025). The Interval-Valued Weighted Feature Ranking Method Based on Aggregation Functions. In 2025 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), Reims, France, 6–10 July 2025 (pp. 1–7). Piscataway, NJ: IEEE. https://doi.org/10.1109/FUZZ62266.2025.11152269

Wojtowicz, A., Mrukowicz, M., Gałka, W., Balicki, K., Rząsa, W., & Bentkowska, U. (2024). Binary Ensemble kNN Based Classifier for Microarray Datasets. Procedia Computer Science, 246, 4411–4420. Knowledge-Based and Intelligent Information & Engineering Systems: Proceedings of the 28th International Conference KES-2024, Seville, Spain, 11–13 September 2024. https://doi.org/10.1016/j.procs.2024.09.291

Bentkowska, U., Gałka, W., Mrukowicz, M., & Wojtowicz, A. (2024). Ensemble Classifier Based on Interval Modeling for Microarray Datasets. Entropy, 26(3), 240. https://doi.org/10.3390/e26030240

Conference abstracts:

Wojtowicz, A., Paja, W., & Bentkowska, U. (2025). A Feature Selection Method Based on Aggregation Functions. In M. Baczyński, P. Grzegorzewski, & K. Miś (Eds.), Abstracts of the 6th International Symposium on Fuzzy Sets (ISFS 2025), Katowice, Poland, 23–25 May 2025 (p. 87). Rzeszów: Polish Society for Fuzzy Sets (POLFUZZ).

Bentkowska, U., Gałka, W., Mrukowicz, M., & Wojtowicz, A. (2023). Multi-class Classification Based on Interval Modelling for Datasets with a Large Number of Conditional Attributes. In Book of Abstracts of the 13th Conference of the European Society for Fuzzy Logic and Technology (EUSFLAT 2023) and the 12th International Summer School on Aggregation Operators (AGOP 2023), Palma de Mallorca, Spain, 4–8 September 2023.

Conference and seminar contributions:

2026

RIMSA Congress 2026 — 4th International Conference of the Rzeszów International Medical Students’ Association, College of Medical Sciences, University of Rzeszów, 23–24 October 2026 (student research conference). Presented work with Ewa Rak and Rafał Rak on conformalized interval type-2 fuzzy monitoring and alert burden using VitalDB.

IWIFSGN 2026 — 24th International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets, Systems Research Institute, Polish Academy of Sciences, Warsaw, 2 October 2026. Co-authored “Single Pass Feature Ranking with Interval Valued Fuzzy Aggregation” with Wiesław Paja, included in the workshop programme (POLFUZZ session Uncertainty and Imprecision in Decision Making).

KES 2026 — 30th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems, Dublin, Ireland, 9–11 September 2026. Co-authored “Interpretable Machine Learning Framework Prototype for FTIR and Raman Spectroscopy in Disease Classification and Spectral Biomarker Discovery”. Wiesław Paja presented the work.

ISD 2026 — 34th International Conference on Information Systems Development, Czech Technical University in Prague, Czechia, 2–4 September 2026. Presented two papers in Track 4 (Data Science and AI for IS Development): on adaptive feature selection and on systemic risk in Central and Eastern European insurance.

Joint computer science seminar of the University of Silesia and the University of Rzeszów (UŚ–UR), spinPLACE, Katowice, 25 June 2026. Poster “Adaptive Interval Feature Selector” with Wiesław Paja.

BAFA Accounting Education SIG Annual Conference & AGM, University of Sheffield, United Kingdom, 20–22 May 2026. Co-authored “The AI Challenge: Prompt Engineering for Accounting and Finance Students” and “AI-Simulated Dialogue for Immersive Learning Across Modules” with Katarzyna Jaskowiec (De Montfort University).

NCN OPUS 27 research seminar “Quantitative methods in the analysis of risk arising from current global threats”, Cracow University of Economics, Kraków, 25 May 2026. Presented “Anatomy of Shock: Machine Learning in Risk and Liquidity Management. Lessons from 2016–2024 and capital forecasts for the CEE insurance sector”, with Ewa Rak.

2025

IWIFSGN 2025 — 23rd International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets, Systems Research Institute, Polish Academy of Sciences, Warsaw, 17 October 2025. Presented “From Research on Feature Importance Aggregation Using Interval-Valued Methods” with Wiesław Paja and Urszula Bentkowska.

FUZZ-IEEE 2025 — IEEE International Conference on Fuzzy Systems, Reims, France, 6–10 July 2025. Co-authored “The Interval-Valued Weighted Feature Ranking Method Based on Aggregation Functions”.

Joint computer science seminar of the University of Silesia and the University of Rzeszów (UŚ–UR), Rzeszów, 28 May 2025. Poster “Feature Ranking Based on Aggregation Functions” with Wiesław Paja and Urszula Bentkowska.

ISFS 2025 — 6th International Symposium on Fuzzy Sets, Katowice, Poland, 23–25 May 2025. Co-authored and delivered the talk “A Feature Selection Method Based on Aggregation Functions” with Wiesław Paja and Urszula Bentkowska (24 May 2025).

2024

Joint computer science seminar of the University of Silesia and the University of Rzeszów (UŚ–UR), Katowice, 28 May 2024. Co-authored “Ensemble kNN-Based Classifier for High-Dimensional Datasets”.

Research projects:

NCN OPUS 27, project no. 2024/53/B/HS4/03258: Identification and measurement of systemic threats to the insurance sector arising from climate risk in the light of sustainable development policy: new tools based on machine learning methods — host institution: Cracow University of Economics; principal investigator: dr hab. Anna Denkowska, prof. UEK. Contribution: machine learning modelling of insurers’ technical provisions and liquidity under macroeconomic shocks.