Scholarlink

<p>A few existing literatures utilized the parameters of heterogeneous autoregressive (HAR) model, which captures the short and long memory information of realized volatility, as the input indicators for forecasting realized volatility. In this work, we leverage the predictive advantages of machine learning models to forecast volatility by integrating HAR model parameters with technical indicators. We apply this hybrid approach to Chinese futures market data to evaluate its performance. Empirical results demonstrate that incorporating HAR parameters significantly enhances forecasting accuracy across six loss functions within the same machine learning framework. Furthermore, the Light Gradient Boosting Machine (LightGBM) exhibits superior forecasting accuracy. Finally, in multi-step forecasting comparisons among different machine learning models, LightGBM demonstrates robust performance. This research provides a more accurate and reliable method for forecasting realized volatility in futures markets.</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

On the Realized Volatility Forecasting Based on Hybrid Model Integrating HAR Model with Machine Learning Method

  • 期刊论文

  • 发表日期：

    2025年5月16日

查看全文
  • Yan Song,
  • Tiantian Yin,
  • Yuping Song

摘要

A few existing literatures utilized the parameters of heterogeneous autoregressive (HAR) model, which captures the short and long memory information of realized volatility, as the input indicators for forecasting realized volatility. In this work, we leverage the predictive advantages of machine learning models to forecast volatility by integrating HAR model parameters with technical indicators. We apply this hybrid approach to Chinese futures market data to evaluate its performance. Empirical results demonstrate that incorporating HAR parameters significantly enhances forecasting accuracy across six loss functions within the same machine learning framework. Furthermore, the Light Gradient Boosting Machine (LightGBM) exhibits superior forecasting accuracy. Finally, in multi-step forecasting comparisons among different machine learning models, LightGBM demonstrates robust performance. This research provides a more accurate and reliable method for forecasting realized volatility in futures markets.