An Adaptive Predictive Arithmetic Based on Phase Space Reconstruction

A new algorithm is proposed for settling the prediction of nonlinear time series. It makes use of the particularities of chaotic attractors and the relativities of unknown signals to reconstruct the phase space. Moreover, the algorithm is optimized and the concept of "near neighbor weighted distance" is imported to differentiate the Euclidian distance of every part of the near neighbor point. Confirming the near neighbor points of the future position adaptively, eliminating the "false near neighbor points", then confirming the parameters of the prediction equation. The prediction experiment of jumping frequency signals is proved that this algorithm has perfect precision and solidity. In summary, this predictive algorithm is better than others.