Abstract
INTERNATIONAL JOURNAL OF ENGINEERING, SCIENCE AND - Volume 8,Issue 4, April 2019
Pages: 98-110
IMPROVING FORECASTS OF GARCH FAMILY MODELS WITH THE ARTIFICIAL NEURAL NETWORKS:AN APPLICATION TO DAILY RETURN VOLATILITY IN MOROCCAN STOCK MARKET.
Moulay Driss ELBOUSTY? Hicham EL BOUSTY?? Lahsen OUBDI??? Salah-ddine KRIT????
Category:Engineering, Science and Mathematics
Abstract:
|
Stock Market volatility has been extensively studied in finance literature. In this paper, we estimate Moroccan Stock Market return volatility by using Single-State GARCH models and Markov Regime Switching GARCH models. We proposed Back Propagation Neural Network algorithms to improve volatility forecasting of GARCH class models. The BPNN is combined with GARCH in such a way prediction of GARCH models is used as input of our Neural Network. Three volatility estimators are used for this purpose: Absolute return, Parkinson and Garman Klass. The forecasting accuracy of the models is examined using Mean Square Errors (MSE). The results indicate the efficiency of the neural network in enhancing the performance of GARCH models. The findings further clarify the superiority of the marriage of MRS-GARCH and EGARCH with neural network over considered models. |
Keywords: GARCH; MRSGARCH; Neural Network; Volatility.