Connectionism ARCH modelprocessing Area of research:Engineering and Information Science and Technology

Networks of neurons

With the existence of volatility, several experts try to explain the reason for volatility. The reason for volatility is not only just existing information, but also several other factors. The ARCH and GARCH models have the advantage of allowing complex time series to be modeled with sufficient parameters, and are for this reason in particular to predict volatility. It is increasingly recognized, however, that the inclusion of non-stationarities in the series is inevitable in order to apply such models to long-term real data. The purpose of this paper was to highlight the utility of non-linear models and conditional heteroscedasticity, which now has powerful analytical and modeling tools based on sound theoretical bases to model stationary time series with nonlinear dynamics. The concept of conditional variance characterizes the models that come to broaden the class of classical models based essentially on a linear dependence structure between a variable at moment t and its past values and those of white noise and its past values. However, ARCH models are problematic when the number of historical data becomes very large in which case conditional variances tend to become negative. Indeed, the problem with ARCH models is that volatility is predicted. But variables tend to be negatively correlated, a phenomenon that ARCH models cannot incorporate because they restrict volatility to be affected only.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.