Scholarlink

Artificial Intelligence (AI) has significantly advanced healthcare, enhancing clinical decision-making and patient care. This study highlights the critical role of Explainable AI (XAI) in preterm birth prediction, emphasizing model interpretability as key to building reliable AI systems. Unlike traditional methods that rely on complex image and signal processing, we propose a novel approach using tabular data from the CDC Birth dataset. To address the “black box” issue, we apply different model-agnostic interpretability techniques like LIME, BayLIME, S-LIME, and SMILE, providing clear, actionable insights into model predictions. By enhancing transparency, XAI fosters greater trust among healthcare professionals, improving clinical outcomes and accountability. This work underscores the importance of explainability in deploying AI for high-stakes healthcare applications. Further, the paper explores the instability of LIME and implements different improvements applied to LIME to make it more stable and consistent with respect to explanation generations. Finally, a comparative analysis of these models is enforced based on a statistical metric, the Jaccard index.

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

Model-Agnostic Interpretability Techniques for Machine Learning-Based Pre-Term Birth Prediction

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jashmin Swain,
  • Krishna Sameer,
  • Siba K. Udgata

摘要

Artificial Intelligence (AI) has significantly advanced healthcare, enhancing clinical decision-making and patient care. This study highlights the critical role of Explainable AI (XAI) in preterm birth prediction, emphasizing model interpretability as key to building reliable AI systems. Unlike traditional methods that rely on complex image and signal processing, we propose a novel approach using tabular data from the CDC Birth dataset. To address the “black box” issue, we apply different model-agnostic interpretability techniques like LIME, BayLIME, S-LIME, and SMILE, providing clear, actionable insights into model predictions. By enhancing transparency, XAI fosters greater trust among healthcare professionals, improving clinical outcomes and accountability. This work underscores the importance of explainability in deploying AI for high-stakes healthcare applications. Further, the paper explores the instability of LIME and implements different improvements applied to LIME to make it more stable and consistent with respect to explanation generations. Finally, a comparative analysis of these models is enforced based on a statistical metric, the Jaccard index.