THE USE OF ARTIFICIAL INTELLIGENCE FOR FORECASTING FINANCIAL MARKET VOLATILITY

Abstract

Introduction. The digitalization of the financial sector, the growth of information flows, and increasing instability of global markets have intensified the importance of forecasting financial market volatility for financial analytics and risk management systems. Traditional econometric models demonstrate limited ability to adapt to nonlinear market dynamics and crisis-related informational changes, which increases the relevance of applying artificial intelligence algorithms in financial forecasting processes.

Purpose. The purpose of the study is to develop and empirically evaluate an adaptive approach to forecasting financial market volatility based on the integration of econometric models and artificial intelligence algorithms.

Materials and methods. The research materials include scientific works of domestic and foreign scholars in the fields of financial analytics, digital technologies, machine learning, and risk management, as well as statistical stock market data. The empirical basis of the study was formed using daily S&P 500 index data for 2020–2025. The study employed methods of theoretical generalization, comparative and economic-statistical analysis, machine learning methods, as well as RMSE and MAE indicators for validating the forecasting accuracy of ARIMA, GARCH, RF, and LSTM models.

Results. The article reveals the economic essence of financial market volatility and identifies the functional capabilities of artificial intelligence models in financial forecasting processes. A comparative assessment of the effectiveness of ARIMA, GARCH, RF, and LSTM models in forecasting market volatility was conducted. The results indicate that the LSTM model demonstrated the highest forecasting accuracy and the lowest RMSE and MAE values compared to classical econometric models. It was established that the use of combined GARCH-LSTM models improves forecasting adaptability under crisis market conditions and reduces forecasting error. An adaptive combined approach to forecasting financial market volatility is proposed, integrating time-series analysis, machine learning algorithms, and an information-background assessment module within a unified system of financial analytics and risk management.

Perspectives. Further research should focus on the development of explainable artificial intelligence models, improvement of adaptive retraining mechanisms for forecasting algorithms, and integration of behavioral analytics and macroeconomic indicators into financial market volatility forecasting systems.