One-step Ahead Prediction of f₀F₂ Based on Phase Space Reconstruction

One-step ahead prediction of critical frequency f0F2 is presented, using samples of about 5856 observations of f0F2 with 15-minute time resolution taken from a station in the low latitude, from March 2005 to April 2005. Based on phase space reconstruction of chaotic time series, and according to the theory that near trajectories have similar characteristics, the k-nearest neighbor method is taken to forecast f0F2 15 minutes ahead. Training method and self-adaptive-select method are attempted to define the neighbor numbers used for prediction, and for these neighbors, averaging method and regression ones are taken to treat them. The forecasting errors for those methods are compared. It is shown that one-step ahead forecasting method based on phase space reconstruction can reach high accuracy. In addition, the method has the advantage of easy to implementing with high processing speed. So it is quite suitable for quasi real-time prediction.