Scholarlink

Heart diseases are one of the leading causes of death among people of all ages across the globe, which raises the need for accurate detection methods. Detecting heart diseases accurately and early is very crucial for decreasing mortality rates. However, the traditional or manual detecting methods are very time-consuming and need experienced manpower. So, the primary goal of a heart disease detection model is to increase the accuracy of classification and to decrease the false positive value. In this paper, we propose a state-of-the-art hybrid model that uses two feature extraction methods named Select-KBest and Feature Elimination with Cross-Validation for extracting the most important features from the dataset with four machine learning models: Logistic Regression (LR), Support Vector Machine (SVM), Naive Bayes (NV) and Random Forest (RF). Then, we employ two ensemble learning methods (Bagging, Boosting) to further increase the accuracy. To solve the data imbalance and overfitting issues, we use the Synthetic Minority Oversampling Method (SMOTE). The proposed hybrid model acquires high accuracy and AUC rates of 99.10% and 99.81%, respectively, with the Logistic Regression and Boosting method, which outperforms the existing works and sets a new standard for detecting heart diseases in real-world scenarios.

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

Enhancing Heart Disease Prediction Model Through SMOTE and Ensemble Learning Techniques

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Md. Zunead Abedin Eidmum,
  • Bakhtiar Muiz,
  • Rakib Hossen,
  • Anichur Rahman

摘要

Heart diseases are one of the leading causes of death among people of all ages across the globe, which raises the need for accurate detection methods. Detecting heart diseases accurately and early is very crucial for decreasing mortality rates. However, the traditional or manual detecting methods are very time-consuming and need experienced manpower. So, the primary goal of a heart disease detection model is to increase the accuracy of classification and to decrease the false positive value. In this paper, we propose a state-of-the-art hybrid model that uses two feature extraction methods named Select-KBest and Feature Elimination with Cross-Validation for extracting the most important features from the dataset with four machine learning models: Logistic Regression (LR), Support Vector Machine (SVM), Naive Bayes (NV) and Random Forest (RF). Then, we employ two ensemble learning methods (Bagging, Boosting) to further increase the accuracy. To solve the data imbalance and overfitting issues, we use the Synthetic Minority Oversampling Method (SMOTE). The proposed hybrid model acquires high accuracy and AUC rates of 99.10% and 99.81%, respectively, with the Logistic Regression and Boosting method, which outperforms the existing works and sets a new standard for detecting heart diseases in real-world scenarios.