Correlation Basis Function network and application to financial market trading

Time series prediction using different kinds of neural networks is widely studied with certain degree of success. We proposed an architecture called Correlation Basis Function (CBF) network to make buying and selling decision on the chaotic financial time series. CBF network, which is functionally similar to Radial Basis Function (RBF) network in forecasting financial time series, can outperform the latter. Our simulations demonstrated that the new network model can bring superior returns in trading some stocks in Hong Kong.