Scholarlink

The increasing adoption of machine learning (ML) models in health care highlights the need for explainable intelligence (XAI) to ensure transparency and trust. Despite the high predictive performance, many ML models lack interpretability. This case study enhances the explainability by implementing an interpretable framework on the Pima Indian Diabetes Dataset (PIDD) using Local Interpretable Model-Agnostic Explanations (LIMEs). We build six ML models, with the random forest classifier giving a 91% accuracy, the highest, and an ROC-AUC score of 0.97. We employed LIME to generate local explanations for individual predictions, revealing feature contributions. We validated interpretability through a local fidelity score, with an average mean absolute error (MAE) of 0.18, indicating high fidelity and reliability. This study showcases the importance of XAI in health care, illustrating how tools like LIME enhance model transparency and trust in AI-driven healthcare solutions.

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

Enhancing Interpretability of Machine Learning Models for Diabetes Prediction Using LIME: A Case Study on the Pima Indian Diabetes Dataset

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Deepthi Angara,
  • Jhansi Yellapu

摘要

The increasing adoption of machine learning (ML) models in health care highlights the need for explainable intelligence (XAI) to ensure transparency and trust. Despite the high predictive performance, many ML models lack interpretability. This case study enhances the explainability by implementing an interpretable framework on the Pima Indian Diabetes Dataset (PIDD) using Local Interpretable Model-Agnostic Explanations (LIMEs). We build six ML models, with the random forest classifier giving a 91% accuracy, the highest, and an ROC-AUC score of 0.97. We employed LIME to generate local explanations for individual predictions, revealing feature contributions. We validated interpretability through a local fidelity score, with an average mean absolute error (MAE) of 0.18, indicating high fidelity and reliability. This study showcases the importance of XAI in health care, illustrating how tools like LIME enhance model transparency and trust in AI-driven healthcare solutions.