Optimizing Hepatitis C Diagnosis: A Comparative Study of Hyperparameters Tuning in Machine Learning Models

Medical science depends heavily on disease classification and prediction to restrain the spread of diseases and early disease detection. Precise disease prediction and classification is a typical application of machine learning (ML) techniques, which is helpful for doctors and other specialists. This study introduces a highly effective and efficient AI-based machine-learning approach to the early diagnosis of the hepatitis C virus. The blood components CHE, CHOL, ALB, ALP, bilirubin (BIL), GGT, PROT, ALT, AST, and CREA are used to identify the virus causing hepatitis C. Because of the large number of patients in LDC countries, the physician sometimes feels tiresome, leading to misdiagnosis. To resolve this issue, we collected a dataset with those features and added the age, sex, and gender of 615 individuals. Later, exploratory data analysis and modern processing techniques were implemented to improve the data’s quality. Subsequently, the three models- Random Forest (RF), logistic regression (LR), and gradient boosting (GB)-were employed to classify them. Grid search hyperparameter tuning was performed to enhance their performance, and the model was evaluated with well-known matrices. The tuned GB and RF classifiers demonstrated the highest accuracy of 0.9774. However, gradient boosting achieves a better AUC. Regarding all matrices, the GB model outperforms in the diagnosis of the hepatitis C virus.