Optimization of Decision Support System Using Dynamic Rough Set and Machine Learning in Complex Dynamic Environments

Resumen

In the era of big data, decision-making in healthcare and other dynamic domains is often chal- lenged by uncertainty, high dimensionality, and evolving datasets. Traditional Decision Support Systems (DSS) struggle to adapt to such complexity, necessitating advanced approaches that balance accuracy with interpretability. This research presents an improved DSS model that incorporates Rough Set Theory (RST) along with Machine Learning (ML) models, specifically Light Gradient Boosting Machine (LightGBM), to improve predictive accuracy and scalability. The approach incorporated routine data preprocessing, rough set–driven feature selection for dimensionality reduction, and classification with RF, XG-Boost, SVM, Lo- gistic Regression, and LightGBM. Experimental performance on four biomedical datasets—hepatic disease, breast cancer, chronic kidney disease, and autism—illustrated that RST-LightGBM uniformly produced bet- ter performance. For example, for classification in breast cancer, LightGBM performed at 97.42% accuracy, 97.05% precision, 97.90% recall, and 97.47% F1-score, much better than current approaches. These out- comes validate the efficiency of the suggested hybrid framework in formulating strong adaptive, flexible, and interpretable DSS for complicated dynamic environments.