Abstract
Forecasting stock market volatility is an important subject in investment and risk management. By improving upon this task, investors can make sounder investment choices in re-balancing their portfolios, leading to greater profits. Artificial Neural Networks (ANNs) have been used extensively in many fields. They have proved to be one of the most effective tools in time series forecasting. The aim of this study is to forecast the South African Volatility Index (SAVI) using a Multilayer Perceptron ANN. Several experiments are conducted where the ANN is trained to forecast (i) the next trading day’s SAVI return and (ii) the movement direction in the next trading day’s SAVI return. The results of the experiments show that the ANN performed better in forecasting the movement direction in the next period’s volatility as opposed to the next period’s expected volatility.
M.Com. (Finance)