Scholarlink

Due to their severity and prevalence, the prediction of chronic diseases (CDs) has become an important area of research, particularly with advances in deep learning. In this paper, we propose a new automated technique for detecting these diseases. A deep network architecture using a parallel unidimensional convolutional neural network (1D-PCNN) is employed to extract deep features. Subsequently, the Support Vector Machine (SVM) technique is applied for CD classification. The uniqueness of our framework lies in the design of the 1D-PCNN, which can learn the deep features of the input layer through parallel convolutional layers. As a result, the deep features of each parallel branch are simultaneously extracted before being combined in the fusion layer. Furthermore, in order to improve the efficiency of the proposed model, the Synthetic Minority Oversampling Technique (SMOTE) is used. This strategy manages class imbalance in CD databases. The suggested model is analysed against standard 1D-CNN, 1D-CNN model combined with conventional machine learning methods and other existing state-of-the-art models. The effectiveness of the suggested method was tested using two known databases the Pima Indian Diabetes Database (PIDD) and the Cleveland Heart Disease Database (CHDD). Results, from these databases indicate that the proposed approach yielded an accuracy rate of 83% and 88% an F score of 73% and 90% and an AUC of 80% and 87% correspondingly. Finally, after applying the SMOTE method, accuracy was improved to 86% and 92%, F-score to 86% and 93%, and AUC to 86% and 92%, respectively, and outperforms other methods.

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

Prediction of Chronic Diseases Using Parallel 1D-CNN Feature Extraction and SVM Classification Based on SMOTE

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Fatma Zohra Tassadit Ait Mesbah,
  • M’hamed Bilal Abidine,
  • Belkacem Fergani

摘要

Due to their severity and prevalence, the prediction of chronic diseases (CDs) has become an important area of research, particularly with advances in deep learning. In this paper, we propose a new automated technique for detecting these diseases. A deep network architecture using a parallel unidimensional convolutional neural network (1D-PCNN) is employed to extract deep features. Subsequently, the Support Vector Machine (SVM) technique is applied for CD classification. The uniqueness of our framework lies in the design of the 1D-PCNN, which can learn the deep features of the input layer through parallel convolutional layers. As a result, the deep features of each parallel branch are simultaneously extracted before being combined in the fusion layer. Furthermore, in order to improve the efficiency of the proposed model, the Synthetic Minority Oversampling Technique (SMOTE) is used. This strategy manages class imbalance in CD databases. The suggested model is analysed against standard 1D-CNN, 1D-CNN model combined with conventional machine learning methods and other existing state-of-the-art models. The effectiveness of the suggested method was tested using two known databases the Pima Indian Diabetes Database (PIDD) and the Cleveland Heart Disease Database (CHDD). Results, from these databases indicate that the proposed approach yielded an accuracy rate of 83% and 88% an F score of 73% and 90% and an AUC of 80% and 87% correspondingly. Finally, after applying the SMOTE method, accuracy was improved to 86% and 92%, F-score to 86% and 93%, and AUC to 86% and 92%, respectively, and outperforms other methods.