A Machine Learning Framework for Non-Invasive Hepatitis C Virus Detection Using Feature Selection

Early and accurate detection of the Hepatitis C Virus (HCV) is essential for timely treatment, especially in low-resource settings where traditional diagnostic methods are invasive and expensive. A machine learning framework for noninvasive multiclass HCV classification using standard laboratory data is presented in this paper. Class imbalance, highdimensionality, and noisy features are three major issues in medical datasets that the framework tackles. We apply a generative AI-based oversampling technique using the Synthetic Data Vault (SDV) to balance the dataset, followed by feature selection methods—Recursive Feature Elimination (RFE), Minimum Redundancy Maximum Relevance (mRMR), and Boruta—to identify optimal feature subsets. Three classifiers—Logistic Regression, Random Forest, and XGBoost—are trained and evaluated using F1-score and ROC-AUC metrics. SHAP-based explainability analysis is also conducted to interpret feature contributions and model predictions. The proposed framework, particularly with RFE and SDV-based synthetic data augmentation, achieves high classification accuracy (96.7%) and improved F1-scores for minority classes, demonstrating its effectiveness for robust and interpretable HCV multiclass detection.