Robust and Explainable Ensemble Based Framework for Liver Disease Classification using Data Balancing and Upsampling

Liver Disease (LD) is a severe health condition impacting over 2 million lives annually worldwide, as reported by the WHO. Factors such as rising alcohol consumption, increased type 2 diabetes cases, genetic predispositions, and various lifestyle influences are expected to heighten LD prevalence further, underscoring the need for a modern, accurate, and interpretable classification system. This paper aims to develop an effective and transparent Machine Learning (ML) framework using ensemble learning models and Explainable AI (XAI) techniques for LD classification. The proposed framework addresses dataset imbalance and size constraints by employing data balancing and upsampling, enabling the ensemble models to learn complex patterns in clinical data. The performance of each model is evaluated, and the best-performing model, Gradient Boosting (GB), is further analyzed using SHAP, LIME, and ELI5 to interpret its feature impact. GB achieved high classification metrics, including accuracy, precision, recall, specificity, and AUC, with Direct Bilirubin, Alkaline Phosphatase, Alanine Aminotransferase, and Age identified as key influential features. This paper successfully presents a reliable and interpretable ML-based framework for LD classification, combining quantitative performance and explainability, making it highly suitable for clinical application.