Scholarlink

<p>The study aims to integrate autoencoder neural networks with denoising techniques, which will advance the field of volatility modeling for financial assets. The data were recreated using an autoencoder neural network while we applied two distinct denoising techniques—wavelet decomposition and variational mode decomposition (VMD)—to the observed financial returns in this study. The study examines the performance of five financial instruments using daily closing prices: the EURUSD, Bitcoin, gold, WTI crude oil, and the SP 500. The models are estimated with both original and denoised returns. The study runs from November 2015 to December 2023. In light of structural changes in financial systems, the results thus support the notion that the application of denoising techniques lowers potential risks while improving model stability and volatility forecast accuracy. This study emphasizes the necessity of integrating neural networks and improved denoising techniques in the domain of risk analysis and financial modeling. Furthermore, improved signal quality has practical implications for financial risk management by improving volatility estimation, facilitating early anomaly detection, and increasing the robustness of forecasting models in dynamically changing market environments.</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

Improving the precision of Markov-switching GARCH models using denoising methods and auto encoder neural networks

  • 期刊论文

  • 发表日期：

    2025年7月9日

查看全文
  • Abdulilah Mubarak

摘要

The study aims to integrate autoencoder neural networks with denoising techniques, which will advance the field of volatility modeling for financial assets. The data were recreated using an autoencoder neural network while we applied two distinct denoising techniques—wavelet decomposition and variational mode decomposition (VMD)—to the observed financial returns in this study. The study examines the performance of five financial instruments using daily closing prices: the EURUSD, Bitcoin, gold, WTI crude oil, and the SP 500. The models are estimated with both original and denoised returns. The study runs from November 2015 to December 2023. In light of structural changes in financial systems, the results thus support the notion that the application of denoising techniques lowers potential risks while improving model stability and volatility forecast accuracy. This study emphasizes the necessity of integrating neural networks and improved denoising techniques in the domain of risk analysis and financial modeling. Furthermore, improved signal quality has practical implications for financial risk management by improving volatility estimation, facilitating early anomaly detection, and increasing the robustness of forecasting models in dynamically changing market environments.