Scholarlink

The leading mortality rates from heart disease worldwide demand advanced predictive models for early disease detection. This could be solved by analyzing complex, high-dimensional medical data, which depends heavily on machine learning because it detects patterns that traditional statistical methods cannot easily identify. Federated learning (FL) was introduced as a machine learning paradigm to train models across distributed devices or servers while maintaining data privacy and security. This study explores the application of FL for heart disease classification using the CVD 2022 dataset. We implement and compare three FL optimization strategies—FedAvg, FedProx, and FedAdam—integrated with various machine learning models, including Logistic Regression, Artificial Neural Networks (ANNs), Decision Trees, Random Forests, and XGBoost. Our results demonstrate that Random Forest achieves the highest accuracy (96.16%) under FedProx.

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 for Heart Disease Prediction: A Privacy-Preserving Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Deepanshu Moyal,
  • Pinky,
  • Vinay Kumar Vats,
  • Karan Verma

摘要

The leading mortality rates from heart disease worldwide demand advanced predictive models for early disease detection. This could be solved by analyzing complex, high-dimensional medical data, which depends heavily on machine learning because it detects patterns that traditional statistical methods cannot easily identify. Federated learning (FL) was introduced as a machine learning paradigm to train models across distributed devices or servers while maintaining data privacy and security. This study explores the application of FL for heart disease classification using the CVD 2022 dataset. We implement and compare three FL optimization strategies—FedAvg, FedProx, and FedAdam—integrated with various machine learning models, including Logistic Regression, Artificial Neural Networks (ANNs), Decision Trees, Random Forests, and XGBoost. Our results demonstrate that Random Forest achieves the highest accuracy (96.16%) under FedProx.