Reducing the number of input variables through symbolic regression

Dejan Brkić; Pavel Praks; Martin Marek; Uroš Ilić; Zoran Stajić; Dejan Brkić; Pavel Praks; Martin Marek; Uroš Ilić; Zoran Stajić

doi:10.3934/era.2025231

Electronic Research Archive

2025, Volume 33, Issue 9: 5158-5178. doi: 10.3934/era.2025231

Previous Article Next Article

Research article Special Issues

Reducing the number of input variables through symbolic regression

1.
Faculty of Electronic Engineering, University of Niš, 18000 Niš, Serbia
2.
IT4Innovations, VSB-Technical University of Ostrava, 708 00 Ostrava, Czech Republic
3.
ENET Centre, VSB – Technical University of Ostrava, 708 00 Ostrava, Czech Republic
4.
Department of Technical Studies, College of Polytechnics Jihlava, 586 01 Jihlava, Czech Republic

Received: 20 December 2024 Revised: 16 June 2025 Accepted: 26 June 2025 Published: 01 September 2025

Symbolic regression, a type of machine learning technique, can efficiently disregard variables that are not significant to the final output, even if they were initially preselected as inputs. Various input parameters are tested in the three examples presented here, where the outputs are modeled using symbolic regression: estimating the middle plasma torch temperature used for waste gasification, the active energy of a solar power plant, and the diameter of a pipe with a known flow and pressure drop through it. Final highly accurate formulas are produced after numerous attempts with lower performances. The process for rejecting the parameters without or with limited influence is automatic and can be performed without human intervention and supervision. The results obtained using symbolic regression are easily interpretable by human experts. This approach shows how to use machine learning-based modeling as an additional tool for sensitivity analysis.
- symbolic regression,
- sensitivity analysis,
- significant input variables,
- waste gasification,
- plasma torch,
- solar plant,
- electricity production,
- pipe diameter
Citation: Dejan Brkić, Pavel Praks, Martin Marek, Uroš Ilić, Zoran Stajić. Reducing the number of input variables through symbolic regression[J]. Electronic Research Archive, 2025, 33(9): 5158-5178. doi: 10.3934/era.2025231

Related Papers:

Abstract

Symbolic regression, a type of machine learning technique, can efficiently disregard variables that are not significant to the final output, even if they were initially preselected as inputs. Various input parameters are tested in the three examples presented here, where the outputs are modeled using symbolic regression: estimating the middle plasma torch temperature used for waste gasification, the active energy of a solar power plant, and the diameter of a pipe with a known flow and pressure drop through it. Final highly accurate formulas are produced after numerous attempts with lower performances. The process for rejecting the parameters without or with limited influence is automatic and can be performed without human intervention and supervision. The results obtained using symbolic regression are easily interpretable by human experts. This approach shows how to use machine learning-based modeling as an additional tool for sensitivity analysis.

References

[1]	D. Angelis, F. Sofos, T. E. Karakasidis, Artificial intelligence in physical sciences: Symbolic regression trends and perspectives, Arch. Comput. Methods Eng., 30 (2023), 3845–3865. https://doi.org/10.1007/s11831-023-09922-z doi: 10.1007/s11831-023-09922-z
[2]	N. Makke, S. Chawla, Interpretable scientific discovery with symbolic regression: A review, Artif. Intell. Rev., 57 (2024), 2. https://doi.org/10.1007/s10462-023-10622-0 doi: 10.1007/s10462-023-10622-0
[3]	M. Marek, D. Brkić, P. Praks, T. Kozubek, J. Frantík, Experimental analysis of magnetic focusing of the plasma arc of a cutting torch, Materials, 18 (2025), 1811. https://doi.org/10.3390/ma18081811 doi: 10.3390/ma18081811
[4]	R. K. Mohanta, D. Kumawat, G. Ravi, Effect of chamber pressure on the output characteristics of a low-pressure DC plasma torch, J. Appl. Phys., 134 (2023), 153302. https://doi.org/10.1063/5.0160624 doi: 10.1063/5.0160624
[5]	E. Sanjaya, A. Abbas, Plasma gasification as an alternative energy-from-waste (EFW) technology for the circular economy: An environmental review, Resour. Conserv. Recycl., 189 (2023), 106730. https://doi.org/10.1016/j.resconrec.2022.106730 doi: 10.1016/j.resconrec.2022.106730
[6]	T. M. Pavlović, I. S. Radonjić, D. D. Milosavljević, L. S. Pantić, A review of concentrating solar power plants in the world and their potential use in Serbia, Renewable Sustainable Energy Rev., 16 (2012), 3891–3902. https://doi.org/10.1016/j.rser.2012.03.042 doi: 10.1016/j.rser.2012.03.042
[7]	D. Brkić, P. Praks, R. Praksová, T. Kozubek, Symbolic regression approaches for the direct calculation of pipe diameter, Axioms, 12 (2023), 850. https://doi.org/10.3390/axioms12090850 doi: 10.3390/axioms12090850
[8]	D. Brkić, Z. Stajić, M. Živković, Sizing pipes without iterative calculus: Solutions for head loss, flow discharge and diameter, in 2023 24th International Carpathian Control Conference (ICCC), IEEE, (2023), 71–76. https://doi.org/10.1109/ICCC57093.2023.10178917
[9]	D. Brkić, Revised friction groups for evaluating hydraulic parameters: Pressure drop, flow, and diameter estimation, J. Mar. Sci. Eng., 12 (2024), 1663. https://doi.org/10.3390/jmse12091663 doi: 10.3390/jmse12091663
[10]	G. Kronberger, B. Burlacu, M. Kommenda, S. M. Winkler, M. Affenzeller, Symbolic Regression, CRC Press, 2024. https://doi.org/10.1201/9781315166407
[11]	R. Dubčáková, Eureqa: Software review, Genet. Program. Evolvable Mach., 12 (2011), 173–178. https://doi.org/10.1007/s10710-010-9124-z doi: 10.1007/s10710-010-9124-z
[12]	F. Llorella, J. A. Cebrián, A. Corbi, A. M. Pérez, Fostering scientific methods in simulations through symbolic regressions, Phys. Educ., 59 (2024), 045010. https://doi.org/10.1088/1361-6552/ad3cad doi: 10.1088/1361-6552/ad3cad
[13]	M. Cranmer, Interpretable machine learning for science with pysr and symbolicregression.jl, preprint, preprint, arXiv: 2305.01582.
[14]	A. Tonda, Review of PySR: High-performance symbolic regression in Python and Julia, Genet. Program. Evolvable Mach., 26 (2025), 7. https://doi.org/10.1007/s10710-024-09503-4 doi: 10.1007/s10710-024-09503-4
[15]	S. M. Udrescu, M. Tegmark, AI Feynman: A physics-inspired method for symbolic regression, Sci. Adv., 6 (2020), eaay2631. https://doi.org/10.1126/sciadv.aay2631 doi: 10.1126/sciadv.aay2631
[16]	K. Wang, T. Shen, J. Wei, J. Liu, W. Hu, An intelligent framework for deriving formulas of aerodynamic forces between high-rise buildings under interference effects using symbolic regression algorithms, J. Build. Eng., 99 (2025), 111614. https://doi.org/10.1016/j.jobe.2024.111614 doi: 10.1016/j.jobe.2024.111614
[17]	P. Kahlmeyer, M. Fischer, J. Giesen, Dimension reduction for symbolic regression, in Proceedings of the AAAI Conference on Artificial Intelligence, AAAI, (2025), 17707–17714. https://doi.org/10.1609/aaai.v39i17.33947
[18]	S. Nguyen-Kuok, The arc plasma torches, in Theory of Low-Temperature Plasma Physics, Springer, (2017), 285–366. https://doi.org/10.1007/978-3-319-43721-7_8
[19]	N. Yu, Y. Yang, R. Jourdain, M. Gourma, A. Bennett, F. Fang, Design and optimization of plasma jet nozzles based on computational fluid dynamics, Int. J. Adv. Manuf. Technol., 108 (2020), 2559–2568. https://doi.org/10.1007/s00170-020-05568-4 doi: 10.1007/s00170-020-05568-4
[20]	A. A. Safronov, V. E. Kuznetsov, O. B. Vasilieva, Y. D. Dudnik, V. N. Shiryaev, AC plasma torches. Arc initiation systems. Design features and applications, Instrum. Exp. Tech., 62 (2019), 193–200. https://doi.org/10.1134/S0020441219020246 doi: 10.1134/S0020441219020246
[21]	M. Skakov, A. Miniyazov, T. Tulenbergenov, I. Sokolov, G. Zhanbolatova, A. Kaiyrbekova, et al., Hydrogen production by methane pyrolysis in the microwave discharge plasma, AIMS Energy, 12 (2024), 548–560. https://doi.org/10.3934/energy.2024026 doi: 10.3934/energy.2024026
[22]	J. Deng, J. Zhang, Q, Zhang, S. Xu, Effects of induction coil parameters of plasma torch on the distribution of temperature and flow fields, Alexandria Eng. J., 60 (2021), 501–510. https://doi.org/10.1016/j.aej.2020.09.022 doi: 10.1016/j.aej.2020.09.022
[23]	G. Piñeiro, S. Perelman, J. P. Guerschman, J. M. Paruelo, How to evaluate models: Observed vs. predicted or predicted vs. observed?, Ecol. Modell., 216 (2008), 316–322. https://doi.org/10.1016/j.ecolmodel.2008.05.006 doi: 10.1016/j.ecolmodel.2008.05.006
[24]	Z. Čorba, D. Milićević, B. Dumnić, B. Popadić, The experiences of the realization of PV power plants after implementation of the prosumers status, J. Process. Energy Agric., 27 (2023), 13–15. https://doi.org/10.5937/jpea27-43506 doi: 10.5937/jpea27-43506
[25]	P. Orzechowski, W. La Cava, J. H. Moore, Where are we now?: A large benchmark study of recent symbolic regression methods, in Proceedings of the Genetic and Evolutionary Computation Conference, Association for Computing Machinery, (2018), 1183–1190. https://doi.org/10.1145/3205455.3205539
[26]	G. S. I. Aldeia, F. O. de França, Interpretability in symbolic regression: A benchmark of explanatory methods using the Feynman data set, Genet. Program. Evolvable Mach., 23 (2022), 309–349. https://doi.org/10.1007/s10710-022-09435-x doi: 10.1007/s10710-022-09435-x
[27]	V. Borisov, T. Leemann, K. Seßler, J. Haug, M. Pawelczyk, G. Kasneci, Deep neural networks and tabular data: A survey, IEEE Trans. Neural Networks Learn. Syst., 35 (2024), 7499–7519. https://doi.org/10.1109/TNNLS.2022.3229161 doi: 10.1109/TNNLS.2022.3229161
[28]	M. Cranmer, A. Sanchez-Gonzalez, P. Battaglia, R. Xu, K. Cranmer, D. Spergel, et al., Discovering symbolic models from deep learning with inductive biases, in Advances in Neural Information Processing Systems, Curran Associates, Inc., 33 (2020), 17429–17442.
[29]	L. Stajić, R. Praksová, D. Brkić, P. Praks, Estimation of global natural gas spot prices using big data and symbolic regression, Resour. Policy, 95 (2024), 105144. https://doi.org/10.1016/j.resourpol.2024.105144 doi: 10.1016/j.resourpol.2024.105144
[30]	P. Praks, M. Lampart, R. Praksová, D. Brkić, T. Kozubek, J. Najser, Selection of appropriate symbolic regression models using statistical and dynamic system criteria: Example of waste gasification, Axioms, 11 (2022), 463. https://doi.org/10.3390/axioms11090463 doi: 10.3390/axioms11090463
[31]	P. Praks, A. Rasmussen, K. O. Lye, J. Martinovič, R. Praksová, F. Watson, et al., Sensitivity analysis of parameters for carbon sequestration: Symbolic regression models based on open porous media reservoir simulators predictions, Heliyon, 10 (2024), e40044. https://doi.org/10.1016/j.heliyon.2024.e40044 doi: 10.1016/j.heliyon.2024.e40044
[32]	M. Schmidt, H. Lipson, Distilling free-form natural laws from experimental data, Science, 324 (2009), 81–85. https://doi.org/10.1126/science.1165893 doi: 10.1126/science.1165893
[33]	C. Wilstrup, J. Kasak, Symbolic regression outperforms other models for small data sets, preprint, arXiv: 2103.15147.
[34]	A. Shmuel, O. Glickman, T. Lazebnik, Machine and deep learning performance in out-of-distribution regressions, Mach. Learn.: Sci. Technol., 5 (2024) 045078. https://doi.org/10.1088/2632-2153/ada221 doi: 10.1088/2632-2153/ada221

era-33-09-231-Supplementary.zip

Reader Comments

Your name:*

Email:*
© 2025 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)