Scholarlink

<p>This study aims to enhance the accuracy of credit risk assessment in financial institutions by addressing the challenges posed by abundant unlabeled data. Traditional support vector machines (SVMs) encounter challenges in semi-supervised learning due to the large number of samples that lack labels. To overcome this limitation, we propose an adaptive fuzzy semi-supervised support vector machine (A-FS<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(^{3}\)</EquationSource> </InlineEquation>VM) approach. A-FS<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(^{3}\)</EquationSource> </InlineEquation>VM utilizes fuzzy membership degrees to dynamically determine the contribution of samples during training, with a focus on increasing the influence of negative samples based on their distribution tightness. The approach is optimized using the adaptive moment estimation (Adam) algorithm. Experimental results on real-world credit risk datasets demonstrate that A-FS<InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(^{3}\)</EquationSource> </InlineEquation>VM achieves superior performance compared to other methods, highlighting its high accuracy and potential for improving the profitability of financial institutions through more precise credit risk assessments.</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

Adaptive fuzzy semi-supervised support vector machine based on sample tightness and its application in credit risk assessment

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • Jing Quan,
  • Xuelian Sun

摘要

This study aims to enhance the accuracy of credit risk assessment in financial institutions by addressing the challenges posed by abundant unlabeled data. Traditional support vector machines (SVMs) encounter challenges in semi-supervised learning due to the large number of samples that lack labels. To overcome this limitation, we propose an adaptive fuzzy semi-supervised support vector machine (A-FS \(^{3}\) VM) approach. A-FS \(^{3}\) VM utilizes fuzzy membership degrees to dynamically determine the contribution of samples during training, with a focus on increasing the influence of negative samples based on their distribution tightness. The approach is optimized using the adaptive moment estimation (Adam) algorithm. Experimental results on real-world credit risk datasets demonstrate that A-FS \(^{3}\) VM achieves superior performance compared to other methods, highlighting its high accuracy and potential for improving the profitability of financial institutions through more precise credit risk assessments.