Title: Explainable Machine Learning System for Predicting Chronic Kidney Disease in High-Risk Cardiovascular Patients

Abstract: As the global population ages, the incidence of Chronic Kidney Disease (CKD) is rising. CKD often remains asymptomatic until advanced stages, which significantly burdens both the healthcare system and patient quality of life. This research developed an explainable machine learning system for predicting CKD in patients with cardiovascular risks, utilizing medical history and laboratory data. The Random Forest model achieved the highest sensitivity of 88.2%. The study introduces a comprehensive explainability framework that extends beyond traditional feature importance methods, incorporating global and local interpretations, bias inspection, biomedical relevance, and safety assessments. Key predictive features identified in global interpretation were the use of diabetic and ACEI/ARB medications, and initial eGFR values. Local interpretation provided model insights through counterfactual explanations, which aligned with other system parts. After conducting a bias inspection, it was found that the initial eGFR values and CKD predictions exhibited some bias, but no significant gender bias was identified. The model's logic, extracted by scoped rules, was confirmed to align with existing medical literature. The safety assessment tested potentially dangerous cases and confirmed that the model behaved safely. This system enhances the explainability, reliability, and accountability of the model, promoting its potential integration into healthcare settings and compliance with upcoming regulatory standards, and showing promise for broader applications in healthcare machine learning.