Combining Chaotic Transformations and Machine Learning Algorithms: Evaluating Explainable Artificial Intelligence Model Performance

Year 2025, Volume: 8 Issue: 1, 45 - 61, 22.06.2025

Cem Özkurt , Eyüp Altuğ Tunç Fadime Zeliha Seyhan Anıl Tunç , Ali Furkan Kamanlı

Abstract

The research presented covers the synthesis of data analysis, machine learning, and explainable artificial intelligence principles. The study investigates chaotic transformations that affect the performance and interpretability of artificial intelligence models in complex systems. Three different chaotic systems were used to transform features in the dataset, including Lorenz, Chen, and Rossler. These transformed datasets were then analyzed using various machine learning algorithms such as Random Forest, Decision Tree and CatBoost. Performance metrics were calculated to evaluate the effectiveness of each combination. Based on these findings, it was observed that the Rossler chaotic system and CatBoost algorithm gave the best results with %99 accuracy, 0.9997 recall and 0.9997 f1 score. The effects of the transformed data on class labels were elucidated using different explainable artificial intelligence models such as ELI5, DALEX and SHAP. Weighted impact analysis outputs were obtained in the range of 3.5 in the SHAP model, 0.035 in the DALEX model and 0.2796 in the ELI5 model. Among the Explainable Artificial Intelligence models, the ELI5 model, which has a more precise range of values, provided the most consistent explanation in our study. Future studies aim to improve the understanding and prediction capabilities of the model by integrating more chaotic systems and machine learning algorithms. Additionally, investigating the robustness of the proposed approach across various datasets and problem domains is anticipated to provide broader applicability and reliability.

Keywords

Chaotic Systems , Machine Learning , Explainable Artificial Intelligence (XAI) , Chaos Theory , Chaotic transformations.

References

• Erik M. Bollt and Joseph D. Skufca. Machine learning for prediction with chaotic data: Applications to chaos synchronization and rogue waves.Chaos: An Interdisciplinary Journal of Nonlinear Science, 28(3):033116, 2018.
• Niklas Bussmann, Paolo Giudici, Dimitri Marinelli, and Jochen Papen- brock. Explainable machine learning in credit risk management. ENGRN:COMPUTER ENGINEERING (TOPIC), 2019. Impact Factor: 4.
• Long Chen, Jiwei Zhang, Tianzhi Sun, and Zhe Xu. Attention mechanism- based long short-term memory network for chaotic time series prediction. IEEE Access, 10:15692–15701, 2022.
• Shoaib Ehsan, Rodrigo Abreu, Usama Anwar, Mehmet E. Celebi, and Mo-hiuddin Ahmed. Model-agnostic meta-explainable methods for visual in- terpretability of deep learning models. IEEE Access, 8:21961–21978, 2020.
• Claudio Gallicchio, Alessio Micheli, and Luca Pedrelli. Deep reservoir com- puting: A critical experimental analysis. Entropy, 20(3):177, 2018.
• Jun Bo Gow and Tamas D. Gedeon. Time series prediction using symbolic regression combined with reinforcement learning. Entropy, 22(2):175, 2020.
• Z. Huang, P. R. Vlachas, and P. Koumoutsakos. Physics-informed recurrent neural networks for turbulent flow prediction. arXiv preprint arXiv:2010.07989, 2020.
• Seth Kaplan, Eric Jang, Leon White, Tanya Berger-Wolf, and Robert L. Grossman. Surrogate modeling in the presence of chaos. Chaos: An Inter- disciplinary Journal of Nonlinear Science, 28(8):085710, 2018.
• Zachary C. Lipton. The mythos of model interpretability. arXiv preprint arXiv:1606.05386, 2016.
• Scott M. Lundberg and Su-In Lee. A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 2017.
• Jaideep Pathak, Brian R. Hunt, Michelle Girvan, Zhixin Lu, and Edward Ott. Model-free prediction of large spatiotemporally chaotic systems from data: A reservoir computing approach. Chaos: An Interdisciplinary Journal of Nonlinear Science, 28(4):041102, 2018.
• B. Ramadevi and Kishore Bingi. Chaotic time series forecasting approaches using machine learning techniques: A review. SYMMETRY, 2022. Impact Factor: 3.15
• Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. Why should I trust you? explaining the predictions of any classifier. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, page 1135–1144, 2016.
• Daniel Smilkov, Nikhil Thorat, Been Kim, Fernanda Vi ́egas, and Martin Wattenberg. Smoothgrad: removing noise by adding noise. arXiv preprint arXiv:1705.07874, 2017.
• Qian Sun, Ying Cheng, Yang Luo, Hong Fan, and Qingqiang Sun. Short-term chaotic time series prediction based on nonlinear attention mechanism. IEEE Access, 9:4877–4885, 2021.
• Chen Wang and Ningde Jin. Robust chaotic time series prediction model based on improved hybrid optimization algorithm. IEEE Access, 9:114924–114937, 2021.
• Xiyu Zhang, Feng Li, Qi Yang, and Tingwen Huang. Sequential prediction of chaotic time series using echo state network optimized by improved cuckoo search algorithm. Entropy, 21(5):444, 2019.
• Jiaming Zhao, Hongjun Cao, and Jianchao Zeng. Predicting chaotic time series using a novel deep learning framework based on a chaotic time series generator. Knowledge-Based Systems, 213:106532, 2021.