Local Functional Coefficient Autoregressive Model for Multistep Prediction of Chaotic Time Series

Algorithm 2

Multistep prediction algorithm.

Input: The relevant data, such as: chaotic time series, the optimal parameters

Output: Results, such as: the prediction values

Scale chaotic time series between ;

Phase space reconstruction based on embedding dimension and the time delay ;

Compute the Euclidean distances between phase points in the reconstructed phase space;

For   %   is unknown data.

  Compute prediction values by using the optimal parameters ;

  Add prediction values to the training set;

  Compute Euclidean distances between the new phase point and others;

End