Scholarlink

<p>Diabetes is a common metabolic disease in which the body has difficulty regulating blood sugar levels. Early and accurate diagnosis can help control the disease to prevent severe complications and improve patient outcomes. This study focused on automatic type-2 diabetes (T2D) diagnosis system. The proposed approach integrates Elastic-Net feature selection with boosting-based classification algorithms. Thus, it is aimed to increase the prediction performance by using the minimum number of features. Additionally, Adaptive Synthetic Sampling (ADASYN) and Synthetic Minority Over-Sampling Technique (SMOTE) were applied to the data to address class imbalance. Unlike conventional approaches, the proposed method simultaneously addresses class imbalance and feature optimization, ensuring both robustness and efficiency. System performance was evaluated by running it on two publicly available diabetes datasets. Experimental results show that the integration of Elastic-Net and boosting algorithms significantly improves diagnostic performance, reaching 100% test accuracy on the first dataset. These findings highlight the novel contribution of this study in combining embedded feature selection and ensemble learning techniques for diabetes diagnosis. The proposed approach offers a promising direction for developing more precise, interpretable, and resource-efficient diagnostic tools in medical data analysis.</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

A Robust Approach To Type-2 Diabetes Diagnosis: Combining Class Imbalance Mitigation, Elastic-Net, and Boosting Algorithms

  • 期刊论文

  • 发表日期：

    2025年12月15日

查看全文
  • Cüneyt Yücelbaş,
  • Şule Yücelbaş

摘要

Diabetes is a common metabolic disease in which the body has difficulty regulating blood sugar levels. Early and accurate diagnosis can help control the disease to prevent severe complications and improve patient outcomes. This study focused on automatic type-2 diabetes (T2D) diagnosis system. The proposed approach integrates Elastic-Net feature selection with boosting-based classification algorithms. Thus, it is aimed to increase the prediction performance by using the minimum number of features. Additionally, Adaptive Synthetic Sampling (ADASYN) and Synthetic Minority Over-Sampling Technique (SMOTE) were applied to the data to address class imbalance. Unlike conventional approaches, the proposed method simultaneously addresses class imbalance and feature optimization, ensuring both robustness and efficiency. System performance was evaluated by running it on two publicly available diabetes datasets. Experimental results show that the integration of Elastic-Net and boosting algorithms significantly improves diagnostic performance, reaching 100% test accuracy on the first dataset. These findings highlight the novel contribution of this study in combining embedded feature selection and ensemble learning techniques for diabetes diagnosis. The proposed approach offers a promising direction for developing more precise, interpretable, and resource-efficient diagnostic tools in medical data analysis.