DEVELOPMENT OF A DECISION SUPPORT SYSTEM USING ARTIFICIAL NEURAL NETWORK AND GENETIC ALGORITHM

Olha Rybak

doi:10.26906/SUNZ.2026.2.176

Автор(и)

Olha Rybak

DOI:

https://doi.org/10.26906/SUNZ.2026.2.176

Ключові слова:

artificial neural network, decision support system, genetic algorithm, artificial intelligence, clustering, evolutionary methods

Анотація

Relevance. Nowadays artificial intelligence technologies are developing rapidly making it possible to automate the most routine component of data processing. AI is based on the computing architecture of a neural network which applies modeling of biological processes that occur in human brains. To improve the structure of neural network for a decision support system and determine its key parameters, such as the number of inputs, the quantity of layers and neurons within each of them, and choosing a training method, this study suggests to use evolutionary methods. The purpose of this research is to investigate the principle of operation of an artificial neural network, whose parameters and structure are determined using genetic algorithm, and to design a decision support system on the basis of the developed model. Research results. Taking into consideration that genetic algorithms software implementation requires a good random number generator and that the basis for the correct functioning of a neural network is the training sample that describes the presented task, it was decided to use the source database of learning materials which can provide parameter values for this purpose. Along with databases, important parts of the developed system are new phenotypes generation block, the block for evaluating them and the neural network training block. The process of a neural network training is preceded by determining a set of training samples and adding noise to them, since the output signals of a well-trained neural network should be insensitive to variations of input values within certain acceptable limits in order to implement monotonic data display. The main criterion when choosing the optimal network architecture is its ability to generalize different types of tasks. Conclusions. Defining parameters of artificial neural network using genetic algorithm allows to simplify the design of its structure and to develop a decision support system on its basis. Experimental results prove that after the training phase is complete the processed data is divided into clusters that correspond to either solution.

Завантажити

Дані для завантаження поки недоступні.

Посилання

1. Hoffmann J., Borgeaud S., Mensch A., Buchatskaya E. et al. (2022). Training Compute-Optimal Large Language Models. In Proceedings of the 36th International Conference on Neural Information Processing Systems (NIPS '22). Curran Associates Inc., Red Hook, NY, USA, Article 2176. – P. 30016–30030. DOI: https://doi.org/10.48550/arXiv.2203.15556

2. Ouyang L., Wu J., Jiang X., Almeida D. et al. (2022). Training language models to follow instructions with human feedback. OpenAI. Journal «Advances in neural information processing systems». Vol. 35. – P. 27730-27744. DOI: https://doi.org/10.48550/arXiv.2203.02155

3. Wei J., Wang X., Schuurmans D., Bosma M. et al. (2022). Chain of Thought Prompting Elicits Reasoning in Large Language Models. In Proceedings of the 36th International Conference on Neural Information Processing Systems (NIPS '22). Curran Associates Inc., Red Hook, NY, USA, Article 1800. – P. 24824–24837. DOI: https://doi.org/10.48550/arXiv.2201.11903

4. Sauer A., Boesel F., Dockhorn T., Blattmann A. et al. (2024). Fast High-Resolution Image Synthesis with Latent Adversarial Diffusion Distillation. In SIGGRAPH Asia 2024 Conference Papers (SA '24). Association for Computing Machinery, New York, NY, USA, Article 106. – P. 1–11. DOI: https://doi.org/10.1145/3680528.3687625

5. Reed S., Żołna K., Parisotto E., Gomez S. et al. (2022). A Generalist Agent. Transactions on Machine Learning Research. – 42 p. DOI: https://doi.org/10.48550/arXiv.2205.06175

6. Fedus W., Zoph B., Shazeer N. (2022). Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity. Journal of Machine Learning Research. Vol. 23, No. 1, Article 120. – P. 5232–5270. DOI: https://doi.org/10.48550/arXiv.2101.03961

7. Zelikman E., Harik G., Shao Y., Jayasiri V. et al. (2024). Quiet-star: Language models can teach themselves to think before speaking. First Conference on Language Modeling. arXiv preprint arXiv:2403.09629. – 26 p. DOI: https://doi.org/10.48550/arXiv.2403.09629

8. Рибак О. (2021). Застосування еволюційних методів оптимізації для вибору режимів шліфування. Технічні науки та технології. Чернігів: НУ «Чернігівська політехніка». № 4(26). – С. 97-105. DOI: https://doi.org/10.25140/2411-5363-2021-4(26)-97-105

9. Bayer H.G., Schwefel H.P., Wegener I. (2002). How to analyse evolutionary algorithms. Theoretical Computer Science, 2002, Vol. 287, Is. 1. – P. 101−130. DOI: https://doi.org/10.1016/S0304-3975(02)00137-8