Exploring the predictability of range-based volatility estimators using recurrent neural networks

Abstract

We investigate the predictability of several range-based stock volatility estimates and compare them with the standard close-to-close estimate, which is most commonly acknowledged as the volatility. The patterns of volatility changes are analysed using long short-term memory recurrent neural networks, which are a state-of-the-art method of sequence learning. We implement the analysis on all current constituents of the Dow Jones Industrial Average index and report averaged evaluation results. We find that the direction of changes in the values of range-based estimates are more predictable than that of the estimate from daily closing values only.