A Novel Prediction Method for ARFIMA Processes

The class of autoregressive fractionally integrated moving average (ARFIMA) model is an important type of long memory processes which are widely used in many fields. In this paper, a novel nonparametric method is proposed to predict ARFIMA processes based on phase space reconstruction theory and multivariate local linear estimator. Moreover, the analytical expression of the mean square error (MSE) of multivariate local linear estimator is deduced in theory. Finally, the computer simulation results show that the proposed method performs better than the conventional methods.