Machine Learning and Ensemble Learning Based Predictive Modeling for Cirrhosis Stage Classification: A Comparative Study

Liver is one of the most critical organs of the human body which performs numerous biological functions. Though liver plays essential role, the liver diseases are widespread and affect overall human health. Among these diseases, Liver cirrhosis is a chronic and progressive disease characterized by the replacement of healthy liver tissue with scar tissue, primarily due to excessive alcohol consumption and Hepatitis B infection. Accurate detection and staging of cirrhosis are essential for early intervention and effective management. This study aims to employ various machine learning models including K-Nearest Neighbors (KNN), Decision Tree, Random Forest, Gradient Boosting, Naïve Bayes, and Support Vector Machine (SVM) to predict the stages of cirrhosis, with a focus on enhancing accuracy through ensemble learning techniques. The scope of this research includes a comparative analysis of these models using two Kaggle datasets: the Liver Cirrhosis Stage Classification dataset and the Cirrhosis Patient Survival Prediction dataset. Our findings indicate that the ensemble model achieved an accuracy of 95.31% on the Liver Cirrhosis Stage Classification dataset, outperforming existing benchmarks. In contrast, the ensemble approach resulted in an accuracy of 56.07% on the Cirrhosis Patient Survival Prediction dataset, highlighting specific limitations related to that dataset. This study underscores the potential of ensemble techniques for predicting the stages of cirrhosis and lays the groundwork for further research in personalized clinical decision support for the management of liver disease.