Scholarlink

<p>Hyperglucagonemia is one of the prevalent diseases which may take several forms such as acute pancreatic damage, hepatitis, renal failure and many more. Timely diagnosis with correct measures is essential for patient care and disease management. Thus, the given paper introduces a hybrid framework comprising phases such as data pre-processing, normalization, establishing relationship between dataset attributes using graph convolutional network (GCN), optimal features selection using Tabu search (TS) algorithm and evaluation of the proposed framework. GCN models the problem as a graph node classification task in which each node acts as a patient. It is followed by optimal selection of instances using TS. TS is used to solve non-convex clustering issues by producing random set of all achievable moves based on current position After this, the framework is tuned and validated based on evaluation metrics such as accuracy (%), precision (%) and recall (%). The results show that the proposed framework produces the highest accuracy, precision and recall as 98.22%, 97.37%, 96.55% respectively, thus outperforming existing studies in terms of performance.</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

TS-GCN: a hybrid intelligent framework combining Tabu search clustering and graph convolutional network for prognosis of hyperglucagonemia

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • Satish Singh Mekale,
  • Maumita Chakraborty,
  • Chiradeep Mukherjee

摘要

Hyperglucagonemia is one of the prevalent diseases which may take several forms such as acute pancreatic damage, hepatitis, renal failure and many more. Timely diagnosis with correct measures is essential for patient care and disease management. Thus, the given paper introduces a hybrid framework comprising phases such as data pre-processing, normalization, establishing relationship between dataset attributes using graph convolutional network (GCN), optimal features selection using Tabu search (TS) algorithm and evaluation of the proposed framework. GCN models the problem as a graph node classification task in which each node acts as a patient. It is followed by optimal selection of instances using TS. TS is used to solve non-convex clustering issues by producing random set of all achievable moves based on current position After this, the framework is tuned and validated based on evaluation metrics such as accuracy (%), precision (%) and recall (%). The results show that the proposed framework produces the highest accuracy, precision and recall as 98.22%, 97.37%, 96.55% respectively, thus outperforming existing studies in terms of performance.