A1613
Title: Forecasting dynamic correlation via a hybrid deep learning: Multivariate DCC GARCH model
Authors: Yasemin Ulu - Easten Michigan University (United States) [presenting]
Abstract: The forecasting performance of the multivariate DCC -GARCH model is compared to that of a hybrid multivariate GARCH deep learning model for the stocks in the BIST30 index. Specifically, a hybrid model based on the recurrent deep neural network (RDNN) and DCC-GARCH models is used. The results are compared to that from a multivariate DCC-GARCH model. The results indicate that the hybrid model that utilizes both DCC-GARCH and deep learning combination performs better than the multivariate GARCH (DCC) model.
