Enlightened prognosis: Hepatitis prediction with an explainable machine learning approach.

Area of Science:

• Hepatology and Computational Medicine
• Artificial Intelligence in Healthcare

Background:

• Hepatitis, a prevalent inflammatory liver condition, poses a significant global health burden.
• Current hepatitis detection methods have limitations, necessitating novel diagnostic strategies.
• Early hepatitis detection is critical for effective patient management and improved outcomes.

Purpose of the Study:

• To investigate the efficacy of traditional machine learning (ML) classifiers for hepatitis infection prediction.
• To compare the performance of logistic regression, SVM, decision trees, random forest, and MLP models.
• To identify the most effective ML approach for early hepatitis detection.

Main Methods:

• Extensive data preprocessing: outlier detection, dataset balancing, and feature engineering.
• Evaluation of ML models using default hyperparameters, GridSearchCV for tuning, and ensemble techniques.
• Performance assessment via accuracy, precision, recall, F1-measure, AUC, and 5-fold cross-validation.

Main Results:

• The Support Vector Machine (SVM) model achieved 99.25% accuracy, 99.27% precision, and 99.24% recall/F1-measure, with a perfect AUC of 1.00.
• Multilayer Perceptron (MLP) and Random Forest models also demonstrated high accuracy (99.00%).
• Explainability analysis identified key features for hepatitis detection, validating model performance.

Conclusions:

• Machine learning, particularly SVM, offers a highly accurate and robust method for hepatitis infection prediction.
• The proposed ML models significantly outperform existing methods in hepatitis detection accuracy and metrics.
• This study highlights the potential of AI in advancing early hepatitis diagnosis and management.