Early Detection and Disease Stage Classification of Viral Hepatitis Using Machine Learning Algorithms

Viral hepatitis represents a major global health concern due to its progression during the early stages with the possibility to evolve into more severe liver diseases. Early and accurate diagnosis is critical to prevent irreversible damage to the liver and to improve health outcomes. This paper presents a comprehensive machine learning–based framework for early detection of viral hepatitis and its stage using clinical laboratory data. Early detection of the disease uses binary classification while its stage classification uses multiclass approach (Blood Donor, Suspect Blood Donor, Hepatitis, Fibrosis and Cirrhosis). A number of machine learning algorithms were evaluated by the authors (Logistic Regression, Support Vector Machine (SVM), Random Forest, Gradient Boosting, Decision Tree, K-Nearest Neighbors and Naïve Bayes) using accuracy, sensitivity, specificity, F1-score, balanced accuracy, ROC-AUC and confusion matrix. Experimental results show that Gradient Boosting algorithm achieved the best performance for early detection in terms of accuracy and ROC-AUC metrics, while Logistic Regression (One-vs-Rest) provided the most reliable performance for stage classification following the same set of metrics. The findings confirm that machine learning algorithms can effectively support early viral hepatitis detection and disease stage classification, offering clinically meaningful decision support for healthcare systems.