Scholarlink

<p>Reliable prediction of option prices under real-world volatility regimes remains a core challenge in economic forecasting. Although the Black–Scholes model is widely used for derivative pricing, its analytical solutions rely on restrictive assumptions that limit its applicability in dynamic market environments. This study proposes a novel application of Least Squares Support Vector Machine (LS-SVM) to solve the pricing problem by reformulating the model within a primal–dual optimization framework. Unlike traditional numerical solvers such as the finite difference method (FDM) and finite element method (FEM), this LS-SVM based formulation leverages kernel-based learning to capture nonlinear pricing dynamics while maintaining explainability. A dual-data framework combining synthetic benchmark data and real market observations is adopted to assess both theoretical robustness and practical relevance. Comparative evaluations reveal that LS-SVM outperforms conventional numerical solvers in terms of stability and computational efficiency, while also surpassing machine learning models including ANN, XGBoost, RF, and classical SVM in predictive accuracy. The approach enhances model interpretability and transparency, offering a scalable and explainable tool for option forecasting under economic uncertainty and supporting data-driven financial decision-making.</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

Kernel-based Explainable Machine Learning for Option Price Prediction in Economic Forecasting under Regime-sensitive Volatility using a Dual Data Approach

  • 期刊论文

  • 发表日期：

    2026年3月27日

查看全文
  • Bhubaneswari Mishra,
  • S. Chakraverty

摘要

Reliable prediction of option prices under real-world volatility regimes remains a core challenge in economic forecasting. Although the Black–Scholes model is widely used for derivative pricing, its analytical solutions rely on restrictive assumptions that limit its applicability in dynamic market environments. This study proposes a novel application of Least Squares Support Vector Machine (LS-SVM) to solve the pricing problem by reformulating the model within a primal–dual optimization framework. Unlike traditional numerical solvers such as the finite difference method (FDM) and finite element method (FEM), this LS-SVM based formulation leverages kernel-based learning to capture nonlinear pricing dynamics while maintaining explainability. A dual-data framework combining synthetic benchmark data and real market observations is adopted to assess both theoretical robustness and practical relevance. Comparative evaluations reveal that LS-SVM outperforms conventional numerical solvers in terms of stability and computational efficiency, while also surpassing machine learning models including ANN, XGBoost, RF, and classical SVM in predictive accuracy. The approach enhances model interpretability and transparency, offering a scalable and explainable tool for option forecasting under economic uncertainty and supporting data-driven financial decision-making.