OPTIMIZATION OF COMPUTER GAME PERFORMANCE USING MACHINE LEARNING METHODS

DOI: https://doi.org/10.32782/2521-6643-2026-1-71.6
Keywords: machine learning, performance optimization, computer games, rendering, neural networks, FPS stabilization, real-time graphics.

Abstract

The article presents a comprehensive method for optimizing the performance of computer games based on machine learning techniques. The proposed approach combines GPU load prediction using deep neural networks (DNN) with adaptive real-time management of rendering parameters, enabling the automatic determination of optimal graphics settings to maintain stable frame rates (FPS), reduce GPU load fluctuations, and preserve high visual quality. A mathematical efficiency model is for- mulated, integrating frame rate, graphics quality, and GPU utilization into an optimization function with weighted coefficients reflecting user priorities. The approach includes adaptive adjustment of graphics parameters based on gradient descent and predictive modeling of computational load, allowing dynamic resource management without developer intervention. To evaluate the effectiveness of the proposed method, comparative experiments were conducted against traditional optimization strategies, including static configuration and linear GPU load prediction. The results demonstrate a 20–30 % increase in FPS, a 10–12 % reduction in average GPU load while maintaining high graphics quality (≈ 98 % of maximum), and a significant reduction in frame rate fluctuations in complex scenes. The prototype implementation, developed in Python, was successfully integrated into modern game engines such as Unity and Unreal Engine, confirming the practical applicability of the method in real-world projects. The scientific novelty of this research lies in the integration of multiple classes of machine learning models into a sin- gle adaptive rendering optimization system, capable of considering both hardware parameters and user behavior in real time. The proposed framework provides a foundation for further development of adaptive ML-based systems in game development, including VR/AR and cloud gaming platforms, as well as for enhancing the efficiency of game production pipelines with high performance and graphics quality requirements. Overall, this method demonstrates the potential to substantially improve real-time performance optimization in interactive entertainment applications while maintaining visual fidelity, providing valuable insights for both academic research and industry practice.

References

1. Завгородній В. В., Завгородня Г. А., Валявська Н. О., Адаменко В. С., Дороговцев Є. В., Несмачний П. В. Метод автоматичної генерації контенту на основі процедурних алгоритмів. Вчені записки Таврійського національного університету ім. В. І. Вернадського. Серія: Технічні науки. 2022. Т. 33, № 1. С. 91–96. DOI: https://doi.org/10.32838/2663-5941/2022.1/15
2. Müller T., Evans A., Schied C., Keller A. Instant Neural Graphics Primitives With a Multiresolution Hash Encoding. ACM Transactions on Graphics. 2022. Vol. 41, No. 4. P. 1–15. DOI: https://doi.org/10.1145/3528223.3530127
3. Завгородній В. В., Завгородня Г. А., Демченко І. В., Крамаренко К. С., Шевченко І. О., Юрченко А. В. Метод створення штучних текстур із заданими параметрами. Вчені записки ТНУ ім. В. І. Вернадського. Серія: Технічні науки. 2022. Т. 33, № 2. С. 86–90. DOI: https://doi.org/10.32838/2663-5941/2022.2/14
4. Завгородній В. В., Завгородня Г. А., Валявська Н. О., Герасименко О. О., Калюжний О. В., Степовий А. В. Пошук аномалій у даних за допомогою машинного навчання. Вчені записки ТНУ ім. В. І. Вернадського. Серія: Технічні науки. 2022. Т. 33, № 3. С. 39–43. URL: https://tech.vernadskyjournals.in.ua/journals/2022/3_2022/6.pdf
5. Xiao L., Nouri S., Chapman M., Fix A., Lanman D., Kaplanyan A. Neural Supersampling for Real Time Rendering. ACM Transactions on Graphics. 2020. Vol. 39, No. 4. Article 142. DOI: https://doi.org/ 10.1145/3386569.3392376
6. Thomas M. M., Liktor G., Peters C., Kim S., Vaidyanathan K., Forbes A. G. Temporally Stable Real Time Joint Neural Denoising and Supersampling. Proceedings of ACM on Computer Graphics and Interactive Techniques HPG. 2022. Vol. 5, No. 3. URL: https://momentsingraphics.de/Publications.html
7. Kurz A., Neff T., Lv Z., Zollhöfer M., Steinberger M. AdaNeRF: Adaptive Sampling for Real-Time Rendering of Neural Radiance Fields. ECCV 2022. URL: https://www.ecva.net/papers/eccv_2022/papers_ECCV/ papers/136770258.pdf
8. Guo Y.-X., Chen G., Dong Y., Tong X. Classifier Guided Temporal Supersampling for Real Time Rendering. Computer Graphics Forum. 2022. Vol. 41, No. 7. P. 237–246. DOI: https://doi.org/10.1111/cgf.14672
9. Kanervisto A., Scheller C., Hautamäki V. Action Space Shaping in Deep Reinforcement Learning. Journal of Machine Learning Research. 2021. DOI: https://doi.org/10.48550/arXiv.2004.00980
10. Souchleris K., Sidiropoulos G.K., Papakostas G.A. Reinforcement Learning in the Game Industry – Review, Prospects and Challenges. Applied Sciences. 2023. Vol. 13, No. 4. Article 2443. DOI: https://doi.org/10.3390/ app13042443
11. Li H., Yang P., Liu W., Yan S., Zhang X., Zhu D. Multi-Agent Reinforcement Learning in Games: Research and Applications. Biomimetics. 2025. Vol. 10, No. 6. Article 375. DOI: https://doi.org/10.3390/biomimetics10060375
12. Wang Q., Zhong Z., Huo Y. et al. State of the Art on Deep Learning-enhanced Rendering Methods. Mach. Intell. Res. 2023. Vol. 20. P. 799–821. DOI: https://doi.org/10.1007/s11633-022-1400-x
13. Jiang M., Li J., Lu Y. et al. D-NeuRA: customizable dynamic neural rendering for human avatars with disentangled pose and appearance. J. King Saud Univ. Comput. Inf. Sci. 2025. Vol. 37. Article 238. DOI: https://doi.org/10.1007/s44443-025-00262-5
14. Yao X., Hu Q., Zhou F., Liu T., Mo Z., Zhu Z., Zhuge Z., Cheng J. SpiNeRF: Direct trained Spiking Neural Networks for Efficient Neural Radiance Field Rendering. Frontiers in Neuroscience. 2025. DOI: https://doi.org/10.3389/fnins.2025.1593580
15. Osemwegie O. Neural Rendering and Deep Learning in Virtual Production: Redefining Cinematic Storytelling, CGI, and Real Time Content Generation. International Journal of Mechanical & Thermal Engineering. 2025. Vol. 6, No. 1. P. 40–52. DOI: https://doi.org/10.22271/27078043.2025.v6.i1a.76
16. Завгородній В. В., Завгородня Г. А., Дроботович К. Є., Тенігін О. В., Шматко М. М. Математичне моделювання у методах формального дослідження. Вчені записки ТНУ ім. В. І. Вернадського. Серія: Технічні науки. 2021. Т. 32, № 6. С. 75–79. DOI: https://doi.org/10.32838/2663-5941/2021.6/12
17. Romanyuk O., Zavalniuk Y., Korobeinikova T., Titova N., Romanyuk S. The Overview of Neural Rendering. Modern Engineering and Innovative Technologies. 2023. Vol. 27, No. 1. P. 129–134. DOI: https://doi.org/10.30890/2567-5273.2023-27-01-060
18. Бешлей М. І., Ковальчук О. В., Андрущак В. С., Бешлей Г. В. Методика оптимізації алгоритмів машинного навчання для вбудованих кіберфізичних систем. Таврійський науковий вісник. Серія: Технічні науки. 2024. DOI: https://doi.org/10.32782/tnv-tech.2024.1.2
19. Piórkowski R., Mantiuk R., Wernikowski M. Learning to predict perceptual visibility of rendering deterioration in computer games. Scientific Reports. 2024. DOI: https://doi.org/10.1038/s41598-024-78254-0
20. Rempe M., Zsolnai Fehér T., Sitzmann V. et al. HyperNeRF: A Higher Dimensional Representation for Topology Changing Neural Radiance Fields. ACM Transactions on Graphics. 2022. DOI: https://doi.org/ 10.1145/3528223.3530144
21. Lombardi S., Simon T., Schwartz G., Zollhöfer M., Sheikh Y., Saragih J. Mixture of Volumetric Primitives for Efficient Neural Rendering. ACM Transactions on Graphics. 2021. Vol. 40, No. 4. Article 59. DOI: https://doi.org/10.48550/arXiv.2103.01954
Published
2026-01-27
How to Cite
Zavhorodnia, G. A., & Zavhorodnii, V. V. (2026). OPTIMIZATION OF COMPUTER GAME PERFORMANCE USING MACHINE LEARNING METHODS. Systems and Technologies, 71(1), 45-51. https://doi.org/10.32782/2521-6643-2026-1-71.6
Issue
Vol. 71 No. 1 (2026): Systems and Technologies
Section
COMPUTER SCIENCES