Scholarlink

Purpose <p>The rapid digitalization of healthcare has enabled data-driven clinical decision support while simultaneously intensifying concerns related to patient privacy, data security, scalability, and model interpretability. Existing privacy-preserving approaches often rely on real patient data or introduce significant computational overhead, limiting their practical adoption. This study aims to develop a privacy-preserving and interpretable healthcare prediction framework that ensures high predictive performance without relying on complex deep learning architectures.</p> Methods <p>A privacy-preserving healthcare prediction framework integrating synthetic patient data with a Fuzzy Broad Learning System (FBLS) is proposed. Synthetic data generation preserves the statistical characteristics of real healthcare records while removing personally identifiable information, ensuring compliance with privacy regulations such as HIPAA and GDPR. Data quality and robustness are enhanced using Dream Optimization Algorithm–Fitness Distance Balance (DOA-FDB) based fuzzy clustering for outlier detection, a Self-Organizing Fuzzy Logic Classifier (SOFLC) for adaptive missing value imputation, and Weighted Fuzzy Entropy (WFE) for interpretable feature selection. The FBLS model combines fuzzy reasoning with broad learning for efficient and explainable classification.</p> Results <p>Experimental results demonstrate strong predictive performance, achieving 95.0% training accuracy, 91.4% testing accuracy, and 93.2% validation accuracy. The proposed framework outperforms existing privacy-preserving and fuzzy learning models while maintaining low computational complexity.</p> Conclusion <p>The findings confirm that the proposed approach effectively balances data utility, privacy preservation, interpretability, and computational efficiency, making it a reliable and ethical solution for secure healthcare analytics.</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

Privacy-preserving healthcare prediction using synthetic patient data and fuzzy broad learning

  • 期刊论文

  • 发表日期：

    2026年3月24日

查看全文
  • Uttam Kumar Lima,
  • Manoranjan Mishra,
  • Suchitra Pattanaik,
  • Sasmita Hota

摘要

Purpose

The rapid digitalization of healthcare has enabled data-driven clinical decision support while simultaneously intensifying concerns related to patient privacy, data security, scalability, and model interpretability. Existing privacy-preserving approaches often rely on real patient data or introduce significant computational overhead, limiting their practical adoption. This study aims to develop a privacy-preserving and interpretable healthcare prediction framework that ensures high predictive performance without relying on complex deep learning architectures.

Methods

A privacy-preserving healthcare prediction framework integrating synthetic patient data with a Fuzzy Broad Learning System (FBLS) is proposed. Synthetic data generation preserves the statistical characteristics of real healthcare records while removing personally identifiable information, ensuring compliance with privacy regulations such as HIPAA and GDPR. Data quality and robustness are enhanced using Dream Optimization Algorithm–Fitness Distance Balance (DOA-FDB) based fuzzy clustering for outlier detection, a Self-Organizing Fuzzy Logic Classifier (SOFLC) for adaptive missing value imputation, and Weighted Fuzzy Entropy (WFE) for interpretable feature selection. The FBLS model combines fuzzy reasoning with broad learning for efficient and explainable classification.

Results

Experimental results demonstrate strong predictive performance, achieving 95.0% training accuracy, 91.4% testing accuracy, and 93.2% validation accuracy. The proposed framework outperforms existing privacy-preserving and fuzzy learning models while maintaining low computational complexity.

Conclusion

The findings confirm that the proposed approach effectively balances data utility, privacy preservation, interpretability, and computational efficiency, making it a reliable and ethical solution for secure healthcare analytics.