 |
Artificial neural network based non-linear transformation of high-frequency returns for volatility forecasting
1 : University of Freiburg
2 : Graduate School of Decision Sciences, University of Konstanz
This paper uses Long Short Term Memory Recurrent Neural Networks to extract information
from the intraday high-frequency returns to forecast daily volatility. Applied to the IBM stock, we
find significant improvements in the forecasting performance of models that use this extracted
information compared to the forecasts of models that omit the extracted information and some of
the most popular alternative models. Furthermore, we find that extracting the information through
Long Short Term Memory Recurrent Neural Networks is superior to two Mixed Data Sampling
alternatives.
