An estimation of nonlinear time series with ARCH errors using neural networks

The design of a neural network-based estimation system for nonlinear economic time series, based on the Volterra model, is presented. We use ECLMS (extended correlation least mean squares) algorithms as the noise cancelling method for nonlinear time series. The validity of the proposed method is demonstrated by estimating a high-order Volterra model with ARCH (autoregressive conditional heteroscedasticity) errors. This algorithm has a good performance in solving nonlinear economic time-series estimation problems.