G.H. Dash, Jr., C.R. Hanumara, and N. Kajiji (USA)
Radial basis function; Neural Networks; Bayesian;Regularization; Volatility; Heterogeneous traders.
This paper compares the modeling performance of three radial basis function (RBF) artificial neural networks (ANN) to the (G)ARCH framework when applied to the high frequency realized volatility patterns of futures options FX contracts. The comparative analytics presented in this research confirm the extant literature on RBF modeling efforts in general and offer new findings that document the ability of the Kajiji-4 Bayesian closed form regularization RBF to produce a statistically smaller modeling error than alternative RBF ANNs when applied to high-frequency realized volatility. Moreover, we find that when subjected to Kajiji-4 examination, the volatility models of global hourly trading of futures options contracts on foreign exchange provide strong evidence to support the existence of heterogeneous traders in the global FX derivatives market.
Important Links:
Go Back
