A systematic method for diagnosis of hepatitis disease using machine learning.

Area of Science:

• Medical Informatics
• Machine Learning Applications in Healthcare
• Hepatitis Disease Diagnosis

Background:

• Hepatitis is a critical global health concern with significant mortality rates.
• Accurate and timely diagnosis of hepatitis is essential for effective treatment and patient outcomes.
• Machine learning (ML) offers promising avenues for improving disease diagnosis through data analysis.

Purpose of the Study:

• To evaluate the performance of distinct machine learning classifiers for hepatitis disease diagnosis.
• To compare classification performance with and without class balancing techniques.
• To identify the most effective ML strategy for systematic hepatitis diagnosis.

Main Methods:

• Utilized the UCI hepatitis dataset for model training and evaluation.
• Assessed Support Vector Machine, Logistic Regression (LR), K-Nearest Neighbor, and Random Forest classifiers.
• Implemented class balancing using the SMOTE (Synthetic Minority Over-sampling Technique) strategy and compared results against models without balancing.

Main Results:

• Class balancing using SMOTE significantly improved the performance of all evaluated classifiers.
• Logistic Regression (LR) combined with SMOTE achieved the highest diagnostic accuracy at 93.18%.
• The study demonstrated the effectiveness of class balancing in enhancing ML model performance for hepatitis diagnosis.

Conclusions:

• Machine learning, particularly Logistic Regression with SMOTE, provides a robust and accurate approach for hepatitis diagnosis.
• Class balancing is a crucial step for optimizing ML model performance when dealing with imbalanced datasets in medical diagnosis.
• The findings support the development of systematic ML-driven strategies for efficient hepatitis disease detection.