Scholarlink

<p>This study presents an explainable ensemble-based recommender system for the early diagnosis of Fatty Liver Disease (FLD). The model integrates SHAP (SHapley Additive Explanations) and LIME (Local Interpretable Model-Agnostic Explanations) to ensure both predictive accuracy and interpretability. Leveraging clinical and demographic features from a real-world liver dataset, the system combines Support Vector Machines, Random Forest, and Gradient Boosting in an ensemble architecture. The integrated explainability framework enables both global and local analysis of feature importance, making the system transparent and clinician-friendly. The proposed model achieved a high accuracy of 94%, along with balanced precision, recall, and AUC scores. Unlike prior studies, our work offers a dual-explanation mechanism embedded within a modular diagnostic recommender, enhancing both trust and usability in clinical settings. This approach bridges the gap between black-box AI models and practical healthcare deployment.</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

Toward transparent diagnosis of fatty liver disease: explainable AI-driven recommender systems using SHAP and LIME

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Md. Zeyauddin,
  • Shafiqul Abidin,
  • Mohmmad Ubaidullah Bokhari,
  • Gaurav Yadav

摘要

This study presents an explainable ensemble-based recommender system for the early diagnosis of Fatty Liver Disease (FLD). The model integrates SHAP (SHapley Additive Explanations) and LIME (Local Interpretable Model-Agnostic Explanations) to ensure both predictive accuracy and interpretability. Leveraging clinical and demographic features from a real-world liver dataset, the system combines Support Vector Machines, Random Forest, and Gradient Boosting in an ensemble architecture. The integrated explainability framework enables both global and local analysis of feature importance, making the system transparent and clinician-friendly. The proposed model achieved a high accuracy of 94%, along with balanced precision, recall, and AUC scores. Unlike prior studies, our work offers a dual-explanation mechanism embedded within a modular diagnostic recommender, enhancing both trust and usability in clinical settings. This approach bridges the gap between black-box AI models and practical healthcare deployment.