Modeling Heterogeneous Risk Behavior of the South African Rand via A Closed-form Radial Basis Function Neural Network

Stylized facts are uncovered for a domestic (U.S.) examination of the South African Rand futures contract (ZAR). In this preliminary study, we model complex volatility patterns by a nonparametric artificial neural network (ANN) that incorporates a performance enhancing closed-form regularization technique. The modeling characteristics revealed by the Kajiji-4 radial basis function (RBF) ANN provides significant new information about the behavior and prediction of the ZAR futures contract. The impact of U.S. based news proves to be an important determinant of volatility prediction. However, the proxy for the U.S. trade weighted dollar index does not prove to be particularly important. Additionally, there is evidence that future studies should continue a focus on the role of one-day lagged behavior. Modeling Hetrogenous Risk.... -2Dash & Kajiji