Improving Hepatitis B outcome prediction with ensemble machine learning: A study on predictive models and interpretability.

Objective: Hepatitis B virus (HBV) is a significant global health threat, responsible for severe liver diseases such as liver failure, cirrhosis, and hepatocellular carcinoma. The burden is especially high in low-income regions, where early diagnosis and treatment are critical for mitigating its impact. This study investigates the effectiveness of various machine learning (ML) techniques in predicting patient outcomes in HBV infection. Methods: The Chi-squared test was used for feature selection to find the most important factors, which were later applied to train and evaluate various ML models. To address the class imbalance in the dataset, the Synthetic Minority Over-sampling Technique (SMOTE) was used to balance the data. SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) were used to improve the models' interpretability. Results: Among individual models, Support Vector Machine (SVM) and Logistic Regression (LR) each achieved an accuracy of 92.5%. By implementing a Voting Classifier that combined SVM and LR, the overall accuracy was improved to 95%. The results showed that higher levels of some risk factors, especially in older patients, greatly raise the risk of death. Conclusion: These insights provide healthcare professionals and policymakers with valuable information to develop predicting better patient outcomes in HBV infection and patient care strategies.