AI-Based Early Disease Prediction System: A Comparative and Explainable Machine Learning Approach for Chronic Kidney and Liver Diseases

This is the authors' abstract. We don't add key points for this paper.

Early prediction of Chronic Kidney Disease (CKD) and liver disease can support timely medical assessment and improve disease management. However, identifying these conditions through clinical and laboratory data can be challenging due to variations in patient characteristics and the complexity of medical attributes. This study proposes an AI-Based Early Disease Prediction System using machine learning techniques to analyse CKD and liver disease datasets. Five algorithms—Logistic Regression, Decision Tree, Random Forest, XGBoost, and LightGBM—will be evaluated to compare their classification performance for both diseases. Data preprocessing techniques will be applied to handle missing values, encode categorical features, and prepare the datasets for model training and testing. Performance will be assessed using accuracy, precision, recall, F1-score, and ROC-AUC, where applicable. In addition, SHapley Additive exPlanations (SHAP) will be used to interpret model predictions and identify features contributing to the output. The proposed framework aims to provide a comparative and interpretable approach to preliminary disease-risk prediction. As a research prototype, it may support healthcare research and early risk assessment but is not intended to replace professional diagnosis. Further testing and independent clinical validation will be necessary before practical clinical use.