Scholarlink

<p>In the current scenario, an increase in disease rate has created a need for the optimization of the healthcare sector for early diagnosis. The optimization of healthcare has been successful in computer-aided disease diagnosis using Machine Learning (ML) techniques. However, the predicted outcomes using ML models are often interpreted as a black box. This leads to the evolution of Explainable Artificial Intelligence (XAI), which clearly describes the reason for the predicted outcome. This paper addresses the limitations of traditional ML techniques, aiming to understand how the black-box ML models can be interpreted through Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP) XAI techniques, specifically designed for structured data. It also compares their effectiveness in interpreting the predictions of Random Forest (RF) and XGBoost (XGB) ensemble ML techniques, using three different structured medical datasets in the healthcare domain, such as breast cancer, diabetes, and stroke datasets. The performance of LIME and SHAP is evaluated in terms of interpretability and transparency by providing clear visual explanations in critical decision-making processes. Among the two ensemble ML techniques analyzed, XGB provides an enhanced accuracy of 99.06% for breast cancer dataset, whereas for diabetes and stroke datasets, RF provides a better accuracy of 76.76% and 94.32% respectively. Furthermore, both LIME and SHAP enhance model transparency and provide valuable insights for healthcare professionals to validate and trust the predicted outcome with a clear visualization.</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

MediXAI System for Predicted Outcome Analysis Using LIME and SHAP for Structured Medical Datasets

  • 期刊论文

  • 发表日期：

    2025年12月17日

查看全文
  • Ramyaa Dhinagaran,
  • Kavitha Srinivasan,
  • S. Mohanavalli,
  • Mohamed Ajwad

摘要

In the current scenario, an increase in disease rate has created a need for the optimization of the healthcare sector for early diagnosis. The optimization of healthcare has been successful in computer-aided disease diagnosis using Machine Learning (ML) techniques. However, the predicted outcomes using ML models are often interpreted as a black box. This leads to the evolution of Explainable Artificial Intelligence (XAI), which clearly describes the reason for the predicted outcome. This paper addresses the limitations of traditional ML techniques, aiming to understand how the black-box ML models can be interpreted through Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP) XAI techniques, specifically designed for structured data. It also compares their effectiveness in interpreting the predictions of Random Forest (RF) and XGBoost (XGB) ensemble ML techniques, using three different structured medical datasets in the healthcare domain, such as breast cancer, diabetes, and stroke datasets. The performance of LIME and SHAP is evaluated in terms of interpretability and transparency by providing clear visual explanations in critical decision-making processes. Among the two ensemble ML techniques analyzed, XGB provides an enhanced accuracy of 99.06% for breast cancer dataset, whereas for diabetes and stroke datasets, RF provides a better accuracy of 76.76% and 94.32% respectively. Furthermore, both LIME and SHAP enhance model transparency and provide valuable insights for healthcare professionals to validate and trust the predicted outcome with a clear visualization.