Predictive Modeling of Hepatitis using Machine Learning Algorithms and Mathematical Formulations

Hepatitis continues to be a major health concern for the world, mostly impacting those who are most at risk, such as infants and pregnant women. This research aims to develop a machine learning model for early hepatitis detection and reduce the number of people affected and killed by the disease. Characterization of Hepatitis B (HBV), Hepatitis C (HCV), and Hepatitis E (HEV) was done using information from a range of demographics, drawn from healthcare facilities and online repositories. Advanced data preprocessing techniques including feature selection, imputation, and normalization were employed to prepare a robust dataset. The machine learning model and the two gradient boosting algorithms of Random Forest, AdaBoost, and XGBoost were trained and tested. The models' performance was assessed using metrics such as Accuracy, Precision, Recall, F1 Score, and Confusion Matrix. XGBoost performed best according to all metrics, giving an accuracy of 0.96, precision of 0.95, recall of 0.94, and F1 score of 0.94. AdaBoost achieved an accuracy of 0.90, precision of 0.90, recall of 0.88, and F1 score of 0.89. Random Forest performed the worst among the others, giving results of 0.85 accuracy, 0.85 precision, 0.84 recall, and 0.84 F1 score. Another check on XGBoost’s performance was done with a ten-fold cross-validation approach. It gave excellent results, reaching an average accuracy of 0.8532 along with precision of 0.8503, recall of 0.8395, and an F1 score of 0.8450. When matched with other models, the proposed model outperformed them and can be applied in a clinical setting.