Scholarlink

<p>Heart disease, also termed cardiovascular disease, is associated with chest pain, sweating and fatigue, which are rapidly increasing nowadays. The worst condition of heart disease leads to death by heart attack. Heart disease can be treated at the earlier stage with proper diagnosis of symptoms, which is a complicated task. In this research, a Deep Recurrent Neural Network (DRNN) that is trained by the proposed White Beluga Whale Optimization (WBWO) is utilized for heart disease prediction. Here, pre-processing for input data is carried out by quantile normalization and missing data imputation. Feature selection is then carried out using the proposed Hybrid Matusita and Ruzicka Similarity Measures (HMRSM). Further, the data augmentation is done by Borderline-SMOTE for selecting the features leading to the prediction of heart disease by DRNN. Moreover, DRNN is trained by WBWO, which is the combination of White Shark Optimization (WSO) and Beluga Whale Optimization (BWO). Furthermore, the performance of HMRSM + WBWO_DRNN for heart disease prediction is evaluated by four metrics, such as accuracy, recall, precision, and F1-score, which attained superior values of 91.9%, 90.6%, 91.7%, and 92.9%, respectively.</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

Hybrid similarity measure-based feature selection and optimally driven deep recurrent neural networks for heart disease prediction

  • 期刊论文

  • 发表日期：

    2026年6月23日

查看全文
  • P. Baby Shamini,
  • B. Jaison

摘要

Heart disease, also termed cardiovascular disease, is associated with chest pain, sweating and fatigue, which are rapidly increasing nowadays. The worst condition of heart disease leads to death by heart attack. Heart disease can be treated at the earlier stage with proper diagnosis of symptoms, which is a complicated task. In this research, a Deep Recurrent Neural Network (DRNN) that is trained by the proposed White Beluga Whale Optimization (WBWO) is utilized for heart disease prediction. Here, pre-processing for input data is carried out by quantile normalization and missing data imputation. Feature selection is then carried out using the proposed Hybrid Matusita and Ruzicka Similarity Measures (HMRSM). Further, the data augmentation is done by Borderline-SMOTE for selecting the features leading to the prediction of heart disease by DRNN. Moreover, DRNN is trained by WBWO, which is the combination of White Shark Optimization (WSO) and Beluga Whale Optimization (BWO). Furthermore, the performance of HMRSM + WBWO_DRNN for heart disease prediction is evaluated by four metrics, such as accuracy, recall, precision, and F1-score, which attained superior values of 91.9%, 90.6%, 91.7%, and 92.9%, respectively.