Scholarlink

<p>Heart disease continues to be one of the leading causes of mortality worldwide, which highlights the immediate need for accurate and interpretable predictive models to support early detection. This work mainly focused a reasonable assessment of various effective machine learning (ML) Algorithms: Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), XG-Boost, Naive Bayes (NB), and K-Nearest Neighbours (KNN) applied to the publicly available UCI heart disease dataset. To address the critical challenge of explainability in clinical decision support systems, Our proposed dual explainability framework combining Local Interpretable Model-Agnostic Explanations (LIME) and Permutation Feature Importance (PFI) was implemented and evaluated using the publicly available UCI Cleveland Heart Disease dataset (n = 303). The framework achieved robust predictive accuracy and consistent interpretability across multiple machine learning models. While the results demonstrate strong internal validation, they are based solely on this dataset, and future work will extend the framework to larger and multi-institutional datasets to ensure broader clinical generalizability. Finally, performance has been evaluated using accuracy, precision, recall, and F1-score. Among all classifiers, the results show that Random Forest and Decision Tree achieved the highest predictive accuracy of 99%. The combined use of LIME and PFI revealed that features such as ST depression (oldpeak), chest pain type, and maximum heart rate (thalach) consistently influenced predictions. This dual-layer interpretability framework enhances the transparency of ML predictions and supports trustworthy AI-driven decision-making in healthcare.</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

Dual explainability framework for heart disease prediction using LIME and permutation feature importance

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月24日

查看全文
  • Zulfikar Ali Ansari,
  • Wasim Khan,
  • Md Shamsul Haque Ansari,
  • Shahin Fatima,
  • Shadab Siddiqui

摘要

Heart disease continues to be one of the leading causes of mortality worldwide, which highlights the immediate need for accurate and interpretable predictive models to support early detection. This work mainly focused a reasonable assessment of various effective machine learning (ML) Algorithms: Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), XG-Boost, Naive Bayes (NB), and K-Nearest Neighbours (KNN) applied to the publicly available UCI heart disease dataset. To address the critical challenge of explainability in clinical decision support systems, Our proposed dual explainability framework combining Local Interpretable Model-Agnostic Explanations (LIME) and Permutation Feature Importance (PFI) was implemented and evaluated using the publicly available UCI Cleveland Heart Disease dataset (n = 303). The framework achieved robust predictive accuracy and consistent interpretability across multiple machine learning models. While the results demonstrate strong internal validation, they are based solely on this dataset, and future work will extend the framework to larger and multi-institutional datasets to ensure broader clinical generalizability. Finally, performance has been evaluated using accuracy, precision, recall, and F1-score. Among all classifiers, the results show that Random Forest and Decision Tree achieved the highest predictive accuracy of 99%. The combined use of LIME and PFI revealed that features such as ST depression (oldpeak), chest pain type, and maximum heart rate (thalach) consistently influenced predictions. This dual-layer interpretability framework enhances the transparency of ML predictions and supports trustworthy AI-driven decision-making in healthcare.