Scholarlink

<p>This study addresses the critical challenge of developing highly accurate yet interpretable machine learning models for liver disease prediction. Using a dataset of 1700 patient records with 10 predictive features, we develop and evaluate multiple machine learning and deep learning approaches. Our ensemble-based Boosting Classifier achieves superior accuracy (90.88%) compared to traditional methods and recent studies. To overcome the “black-box” limitation of complex models, we integrate SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) techniques, providing both global and individual-level insights into model decisions. The interpretability analysis reveals Liver Function Test, Alcohol Consumption, and Age as the most significant predictors, aligning with clinical knowledge. This integrated framework bridges the gap between high predictive performance and clinical explainability, facilitating the adoption of AI-driven diagnostic tools in healthcare settings. Our approach demonstrates significant improvement over existing methods while maintaining transparency in decision-making processes essential for clinical implementation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning and Explainable AI for Liver Disease Prediction: An Integrated Interpretability Framework

  • 期刊论文

  • 发表日期：

    2025年6月24日

查看全文
  • Niaz Ashraf Khan,
  • Md. Ferdous Bin Hafiz,
  • Md. Aktaruzzaman Pramanik,
  • Sohrab Hossain,
  • Shohag Barman,
  • Noman Hossain

摘要

This study addresses the critical challenge of developing highly accurate yet interpretable machine learning models for liver disease prediction. Using a dataset of 1700 patient records with 10 predictive features, we develop and evaluate multiple machine learning and deep learning approaches. Our ensemble-based Boosting Classifier achieves superior accuracy (90.88%) compared to traditional methods and recent studies. To overcome the “black-box” limitation of complex models, we integrate SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) techniques, providing both global and individual-level insights into model decisions. The interpretability analysis reveals Liver Function Test, Alcohol Consumption, and Age as the most significant predictors, aligning with clinical knowledge. This integrated framework bridges the gap between high predictive performance and clinical explainability, facilitating the adoption of AI-driven diagnostic tools in healthcare settings. Our approach demonstrates significant improvement over existing methods while maintaining transparency in decision-making processes essential for clinical implementation.