Forecasting The Volatility of Bist 100 Index Return with Linear and Nonlinear Time Series Models

The successful modeling and forecasting of volatility, which is the most important element of risk indicators, minimizes financial uncertainties. Classical volatility models are insufficient to forecast structural changes in economic variables. In fact, with the recent increase in the number of artificial neural network studies, hybrid models with the combined advantages of multiple model structures have gained importance. The aim of this study is to demonstrate that hybrid models are more reliable and consistent models in forecasting volatility in variables. For this purpose, the return volatility of the Borsa Istanbul 100 index was modeled, and forecasting performance results were compared with hybrid models. According to the findings, the best forecasting performance was achieved with hybrid structures containing the exponential GARCH-Artificial Neural Networks (MSEGARCH-ANN) combination. It can be said that hybrid models are superior in the risk analysis of volatile financial instruments and in the estimation of macroeconomic variables in general.