Scholarlink

<p>In studies related to the prediction on corporate financial distress/risk, achieving efficient and accurate risk prediction for imbalanced datasets is an important topic. This paper explores methods for efficiently training imbalanced datasets while ensuring that the trained models perform well in real-world market predictions. It also compares the performance of different models (Logistic, DNN, CNN, LSTM, Transformer, XGBoost, SVM) in real market data. Unlike previous research, this paper introduces the mathematical theory of the Signature algorithm and verifies that this algorithm significantly improves prediction performance across all models. These conclusions will bring new perspectives to future research on the predictions of imbalanced datasets and the combined application of mathematical theories and deep learning.</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

Financial distress prediction using signatures: evidence from Chinese listed firms

  • 期刊论文

  • 发表日期：

    2025年12月8日

查看全文
  • Jiaqi Kuang,
  • Zihao Guo,
  • Jinghan Wang,
  • Yezhen Wang,
  • Kaiwen Zhang

摘要

In studies related to the prediction on corporate financial distress/risk, achieving efficient and accurate risk prediction for imbalanced datasets is an important topic. This paper explores methods for efficiently training imbalanced datasets while ensuring that the trained models perform well in real-world market predictions. It also compares the performance of different models (Logistic, DNN, CNN, LSTM, Transformer, XGBoost, SVM) in real market data. Unlike previous research, this paper introduces the mathematical theory of the Signature algorithm and verifies that this algorithm significantly improves prediction performance across all models. These conclusions will bring new perspectives to future research on the predictions of imbalanced datasets and the combined application of mathematical theories and deep learning.