Scholarlink

<p>Cardiac disease poses a significant global health challenge, underscoring the critical need for precise diagnostic and treatment approaches. Enhancing patient outcomes and reducing the burden of heart-related ailments hinges on efficient identification of cardiac issues through accurate diagnostic tools and reliable differentiation between affected and unaffected individuals. In addressing these challenges, our study introduces an innovative ensemble method employing a majority voting scheme. This approach aims to improve the accuracy of identifying cardiac issues by leveraging the strengths of three machine-learning classification algorithms: Random Forest (RF), K-Nearest Neighbor (KNN), and Support Vector Machine (SVM). Evaluation utilizes the NHMP dataset from the Department of Cardiology at Nishter Hospital in Multan, Punjab, Pakistan. Various feature selection algorithms, including ANOVA, Boruta with XGBoost, Boruta with Random Forest, Chi Square, and LASSO, are integrated into the ensemble model to optimize performance metrics, primarily focusing on accuracy. Two iterations of the ensemble model are assessed: one employing a majority voting approach (VH) and the other utilizing a soft voting approach (VS). Throughout the evaluation, the ensemble model consistently demonstrates high accuracy levels, ranging from 0.97 to 0.996. Notably, the highest accuracy of 0.996 is achieved when utilizing the LASSO feature selection algorithm with the majority voting strategy (VH). Comparison between the majority voting and soft voting strategies reveals that the majority voting strategy generally yields slightly superior accuracy values across all feature selection algorithms, although with minimal differences. The proposed ensemble method integrating RF, KNN, and SVM classifiers with a majority voting scheme proves effective in enhancing the accuracy of cardiac issue identification. This approach offers promising potential for improving patient care through more precise diagnostic outcomes. By addressing the complexities associated with cardiac disease diagnosis, our study underscores the importance of robust ensemble methods in advancing healthcare solutions aimed at mitigating the impact of cardiac ailments globally.</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

Boosting Machine Learning Accuracy for Cardiac Disease Prediction: The Role of Advanced Feature Engineering and Model Optimization

  • 期刊论文

  • 发表日期：

    2025年9月24日

查看全文
  • Omair Bilal,
  • Arash Hekmat,
  • Inzamam Shahzad,
  • Asif Raza,
  • Saif Ur Rehman Khan

摘要

Cardiac disease poses a significant global health challenge, underscoring the critical need for precise diagnostic and treatment approaches. Enhancing patient outcomes and reducing the burden of heart-related ailments hinges on efficient identification of cardiac issues through accurate diagnostic tools and reliable differentiation between affected and unaffected individuals. In addressing these challenges, our study introduces an innovative ensemble method employing a majority voting scheme. This approach aims to improve the accuracy of identifying cardiac issues by leveraging the strengths of three machine-learning classification algorithms: Random Forest (RF), K-Nearest Neighbor (KNN), and Support Vector Machine (SVM). Evaluation utilizes the NHMP dataset from the Department of Cardiology at Nishter Hospital in Multan, Punjab, Pakistan. Various feature selection algorithms, including ANOVA, Boruta with XGBoost, Boruta with Random Forest, Chi Square, and LASSO, are integrated into the ensemble model to optimize performance metrics, primarily focusing on accuracy. Two iterations of the ensemble model are assessed: one employing a majority voting approach (VH) and the other utilizing a soft voting approach (VS). Throughout the evaluation, the ensemble model consistently demonstrates high accuracy levels, ranging from 0.97 to 0.996. Notably, the highest accuracy of 0.996 is achieved when utilizing the LASSO feature selection algorithm with the majority voting strategy (VH). Comparison between the majority voting and soft voting strategies reveals that the majority voting strategy generally yields slightly superior accuracy values across all feature selection algorithms, although with minimal differences. The proposed ensemble method integrating RF, KNN, and SVM classifiers with a majority voting scheme proves effective in enhancing the accuracy of cardiac issue identification. This approach offers promising potential for improving patient care through more precise diagnostic outcomes. By addressing the complexities associated with cardiac disease diagnosis, our study underscores the importance of robust ensemble methods in advancing healthcare solutions aimed at mitigating the impact of cardiac ailments globally.