Scholarlink

<p>Energy futures markets are notoriously volatile, complicating risk management and investment strategies, particularly during geopolitical crises. To improve uncertainty forecasting under such conditions, we integrate the Multivariate Factor Stochastic Volatility (MFSV) model with Bayesian Optimization (BO), tuning hyperparameters via the Tree-Structured Parzen Estimator (TPE) and evaluating performance using the Continuous Ranked Probability Score (CRPS). In 50-day-ahead forecasts for Brent, West Texas Intermediate (WTI), Dubai Crude, Gasoline, and Heating Oil (HO), the model produces stable covariance predictions. However, sensitivity analysis shows that only some hyperparameters exert meaningful marginal effects, with inadequate tuning leading to wider prediction intervals despite well-captured volatility dynamics.</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

Bayesian-tuned Multivariate Factor Stochastic Volatility Model: Forecasting Amid Geopolitical Crisis

  • 期刊论文

  • 发表日期：

    2025年12月20日

查看全文
  • Laura Manteigas

摘要

Energy futures markets are notoriously volatile, complicating risk management and investment strategies, particularly during geopolitical crises. To improve uncertainty forecasting under such conditions, we integrate the Multivariate Factor Stochastic Volatility (MFSV) model with Bayesian Optimization (BO), tuning hyperparameters via the Tree-Structured Parzen Estimator (TPE) and evaluating performance using the Continuous Ranked Probability Score (CRPS). In 50-day-ahead forecasts for Brent, West Texas Intermediate (WTI), Dubai Crude, Gasoline, and Heating Oil (HO), the model produces stable covariance predictions. However, sensitivity analysis shows that only some hyperparameters exert meaningful marginal effects, with inadequate tuning leading to wider prediction intervals despite well-captured volatility dynamics.