This paper analyzes the performance of temporal fusion transformers in forecasting realized volatilities of stocks listed in the S&P 500 in volatile periods by comparing the predictions with those of state-of-the-art machine learning methods as well as GARCH models. The models are trained on...