Scholarlink

<p>The effective identification of concealed characteristics within credit risk information is crucial for mitigating risks in small enterprises and sustaining stable profits in financial institutions. This study develops a two-stage convolutional neural network hybrid model to elucidate the nonlinear and obscured correlations between data features and default status, significantly enhancing the precision of default predictions. Moreover, this study assesses the profitability based on real data samples from one regional commercial bank in China. The findings demonstrate that our model exhibits superior discriminatory accuracy with high-dimensional, multi-source data. It also improves profit margins by 5–28% compared to other models, by effectively balancing the costs associated with Type I and Type II errors.</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

Two-stage convolutional neural network hybrid model for credit risk evaluation of small enterprises

  • 期刊论文

  • 发表日期：

    2026年1月3日

查看全文
  • Bin Meng,
  • Hongpeng Wang,
  • Zhipeng Zhang,
  • Julien Chevallier

摘要

The effective identification of concealed characteristics within credit risk information is crucial for mitigating risks in small enterprises and sustaining stable profits in financial institutions. This study develops a two-stage convolutional neural network hybrid model to elucidate the nonlinear and obscured correlations between data features and default status, significantly enhancing the precision of default predictions. Moreover, this study assesses the profitability based on real data samples from one regional commercial bank in China. The findings demonstrate that our model exhibits superior discriminatory accuracy with high-dimensional, multi-source data. It also improves profit margins by 5–28% compared to other models, by effectively balancing the costs associated with Type I and Type II errors.