Scholarlink

Cardiovascular diseases (CVDs) continue to be a significant global health challenge, underscoring the need for accurate risk prediction models to improve early detection and prevention. In this work two datasets from different sources are combined; the necessary features from the two sources were used to predict instances of heart disease depending on cholesterol, blood pressure, fasting blood sugar, and many other factors. The validation set comprised 1,918 instances and were trained to evaluate a number of machine learning algorithms such as Logistic Regression, Naive Bayes, Elastic Net (E-Net), Bagging ensemble model, Sparse SVM and Gaussian process regression. The main contribution of the work is to understand how the machine learning models learn when the data is very limited, say two samples (one sample from each class). To check the robustness, the analysis was carried out on two-samples training sets from 500 distinct combinations of heart diseases and normal cases.

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

Learning with Limited Data: A Machine Learning Approach for Heart Disease Prediction

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • V. N. Manjunath Aradhya,
  • P. Shiva Prasad,
  • C. S. Chaithra

摘要

Cardiovascular diseases (CVDs) continue to be a significant global health challenge, underscoring the need for accurate risk prediction models to improve early detection and prevention. In this work two datasets from different sources are combined; the necessary features from the two sources were used to predict instances of heart disease depending on cholesterol, blood pressure, fasting blood sugar, and many other factors. The validation set comprised 1,918 instances and were trained to evaluate a number of machine learning algorithms such as Logistic Regression, Naive Bayes, Elastic Net (E-Net), Bagging ensemble model, Sparse SVM and Gaussian process regression. The main contribution of the work is to understand how the machine learning models learn when the data is very limited, say two samples (one sample from each class). To check the robustness, the analysis was carried out on two-samples training sets from 500 distinct combinations of heart diseases and normal cases.