A Dynamic Cost-sensitive and Explainable Ensemble Framework for Early Detection of Polycystic Ovary Syndrome

Abstract

PCOS  is a prevalent endocrine disorder that affects women of reproductive age. It often leads to infertility, metabolic dysfunction, and psychological problems. Common problems with standard machine learning models for PCOS diagnosis include class imbalance, static design, and lack of model transparency, all of which are important considerations in clinical treatment. In order to solve these challenges this study proposes a dynamic, cost-sensitive, and explainable ensemble framework that integrates KNN, SVM, and EBM using the META-DES dynamic selection mechanism. The framework enhances transparency through LIME-based local explanations, adapts to patient-specific data patterns, and reduces the risk of misclassifying high-risk cases. Comparative evaluation with state-of-the-art models demonstrates superior diagnostic accuracy, reliability, and clinical interpretability. To improve clinician trust, the suggested method provides localized, understandable explanations, adjusts to patient-specific data, and lessens the impact of incorrectly classifying high-risk situations. The framework's capacity to provide enhanced diagnostic reliability and tailored decision assistance is demonstrated through comparison with cutting-edge models. This study advances the application of explainable AI in healthcare by offering a scalable, interpretable, and reliable method for early PCOS detection.