Modeling and Forecasting Stock Market Volatility by Gaussian Processes based on GARCH, EGARCH and GJR Models

Probabilistic methods called Gaussian processes have been successfully shown as a powerful tool for modeling time series data and prediction problem as they are a Bayesian approach with kernel based learning: In this paper, the Gaussian processes are applied to model and predict financial volatility based on GARCH, EGARCH and GJR. Five different kernels are used to train each of the proposed volatility models. More precisely, the experimental results show that, the nonlinear hybrid models can capture well symmetric and asymmetric effects of news on volatility and yields better predictive performance than the classic GARCH, EGARCH and GJR approaches.