A0965
Title: Stock market volatility prediction using a hybrid XGBoost and dynamic conditional correlation model
Authors: Bernard Omolo - University of South Carolina - Upstate (United States) [presenting]
Anas Mohammed - University of KwaZulu-Natal (South Africa)
Abstract: Stock market volatility is a critical measure of financial risk, influencing investment decisions, trading strategies, and risk management practices. Accurate volatility forecasting is, therefore, essential for optimizing financial outcomes. Data from five global stock markets, representing both emerging and developed economies, are leveraged and sourced from Yahoo Finance to evaluate hybrid models that integrate econometric and machine learning techniques. Novel weighted hybrid models and a DCC-XGBoost framework are proposed, combining the dynamic conditional correlation (DCC) model with XGBoost to enhance volatility prediction. The models are rigorously evaluated using RMSE, MAE, and R metrics. Results show that the DCC-XGBoost hybrid model outperforms all benchmarks, achieving superior accuracy in capturing volatility dynamics. This demonstrates the efficacy of blending econometric methods (for structured volatility patterns) with machine learning (for nonlinear relationships), offering a robust, data-driven solution for financial forecasting. The findings highlight the potential of hybrid modeling to balance interpretability and predictive power with practical implications for portfolio optimization and derivative pricing.
