INTEGRATING THE STOCHASTIC VOLATILITY MODEL WITH ARTIFICIAL INTELLIGENCE (TRANSFORMER MODEL) FOR ANALYZING AND FORECASTING THE VOLATILITY OF S&P 500 INDEX STOCK PRICES

Abstract

This study aimed to develop a hybrid model combining the Stochastic Volatility (SV) model with Artificial Intelligence (Transformer Model) to analyze and predict stock price volatility, applied to the S&P500 index during the period from January 3, 2001 to November 30, 2024. The objective was to enhance the accuracy of financial returns predictions by integrating the conditional volatility outputs from the SV model into the Transformer model. Both models were evaluated using time-series data and performance metrics, including MSE, RMSE, and MAE, to measure prediction accuracy. The implementation was carried out in Python leveraging its relevant libraries.

The results demonstrated that the hybrid model outperformed the simple Transformer model, as performance metrics values showed a significant decrease. This indicates that incorporating SV outputs as an additional source of information improved the Transformer model's ability to capture temporal patterns, thereby reducing significant predictions errors.