Scholarlink

Modeling interest rate dynamics is essential for financial decision-making, risk evaluation, and the development of monetary policy. Conventional econometric models frequently fail to encapsulate the intricate, nonlinear dynamics characteristic of interest rate variations. This research utilizes deep learning methodologies: Bidirectional Long Short-Term Memory (BiLSTM) & Bidirectional Gated Recurrent Unit (BiGRU), to improve the precision of interest rate predictions. We utilize multiple technical indicators to derive important aspects that reflect trends, momentum, and volatility in interest rate fluctuations. A historical data of the US Treasury 10-Year Yield has been analyzed for modeling interest rate dynamics. Model evaluation is performed with essential performance indicators, such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and \(R^2\) score. Furthermore, paired t-tests are conducted to evaluate the statistical significance of predictive discrepancies between models. The results demonstrate that both BiLSTM and BiGRU effectively capture temporal dependencies than standard LSTM and GRU in interest rate movements, with BiGRU exhibiting slightly superior predictive accuracy.

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

Advancing Yield Curve Analysis: Integrating Deep Learning Models with Technical Indicators

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Indu Rani,
  • Chandan Kumar Verma,
  • Akanksha Sharma

摘要

Modeling interest rate dynamics is essential for financial decision-making, risk evaluation, and the development of monetary policy. Conventional econometric models frequently fail to encapsulate the intricate, nonlinear dynamics characteristic of interest rate variations. This research utilizes deep learning methodologies: Bidirectional Long Short-Term Memory (BiLSTM) & Bidirectional Gated Recurrent Unit (BiGRU), to improve the precision of interest rate predictions. We utilize multiple technical indicators to derive important aspects that reflect trends, momentum, and volatility in interest rate fluctuations. A historical data of the US Treasury 10-Year Yield has been analyzed for modeling interest rate dynamics. Model evaluation is performed with essential performance indicators, such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and \(R^2\) score. Furthermore, paired t-tests are conducted to evaluate the statistical significance of predictive discrepancies between models. The results demonstrate that both BiLSTM and BiGRU effectively capture temporal dependencies than standard LSTM and GRU in interest rate movements, with BiGRU exhibiting slightly superior predictive accuracy.