Scholarlink

<p>An increasing number of connected devices have emerged in the healthcare sector, producing enormous amounts of personal data and thereby creating significant data privacy, computation workload, and real-time decision-making challenges. This paper addresses these critical issues by proposing a novel FL-DRL framework that leverages Federated Learning (FL) and Deep Reinforcement Learning (DRL) for privacy-preserving, scalable, and real-time healthcare analytics. In our approach, FL ensures that patient data remains localised and secure at the edge, while DRL enables adaptive, real-time decision support for prognostic clinical tasks such as diagnosis and treatment recommendations. The proposed solution is designed to be scalable for large healthcare systems and strictly adheres to privacy requirements. To evaluate our approach, we implemented comprehensive data preprocessing steps, including data cleaning, feature extraction, and normalisation, on the MIMIC-III dataset. Extensive statistical analysis demonstrates that our FL-DRL model achieves higher performance, specifically, a 10% improvement in prediction accuracy and 15% reduction in computational time compared to conventional models, as well as superior precision, recall, and F1-scores. We also discuss the interpretability of our model’s decisions and address potential biases in the dataset that may impact generalizability. Consequently, these results highlight the promise of the proposed framework in effectively addressing data privacy, scalability, and real-time decision-making challenges in intelligent 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

Federated learning and deep reinforcement networks for privacy-preserving real-time data analytics in smart healthcare systems

  • 期刊论文

  • 发表日期：

    2025年10月17日

查看全文
  • Veeramalai Sankaradass,
  • V. K. Manindra Manish

摘要

An increasing number of connected devices have emerged in the healthcare sector, producing enormous amounts of personal data and thereby creating significant data privacy, computation workload, and real-time decision-making challenges. This paper addresses these critical issues by proposing a novel FL-DRL framework that leverages Federated Learning (FL) and Deep Reinforcement Learning (DRL) for privacy-preserving, scalable, and real-time healthcare analytics. In our approach, FL ensures that patient data remains localised and secure at the edge, while DRL enables adaptive, real-time decision support for prognostic clinical tasks such as diagnosis and treatment recommendations. The proposed solution is designed to be scalable for large healthcare systems and strictly adheres to privacy requirements. To evaluate our approach, we implemented comprehensive data preprocessing steps, including data cleaning, feature extraction, and normalisation, on the MIMIC-III dataset. Extensive statistical analysis demonstrates that our FL-DRL model achieves higher performance, specifically, a 10% improvement in prediction accuracy and 15% reduction in computational time compared to conventional models, as well as superior precision, recall, and F1-scores. We also discuss the interpretability of our model’s decisions and address potential biases in the dataset that may impact generalizability. Consequently, these results highlight the promise of the proposed framework in effectively addressing data privacy, scalability, and real-time decision-making challenges in intelligent healthcare analytics.