Abstract
This paper comes up with the adaptive prediction method of Volterra series method in chaotic time series based on matrix factorization method. Taking the monthly runoff of the Huaxian Hydrological Station as example, based on phase-space reconstruction, it identifies the chaotic characteristic through correlation dimension and Lyapunov index. Based on Volterra adaptive filter model, use matrix factorization to solve the equation, which avoid the local optimum problem caused by selecting initial value in Normalized Least Mean Square (NLMS), and at the same time, obtain the global optimal Volterra filter coefficient. According to the Numerical experiments, the prediction performance based on matrix factorization method has no problems in selecting the initial value and can achieve highly-accurate prediction. The conclusions are as follows: (1) The monthly runoff in Huaxian has chaotic characteristic. It’s embedding dimension and optimal embedding dimension is 3; (2) Matrix factorization avoids square root operation and improves the speed; (3) Avoiding selecting the initial value improves the prediction accuracy of the optimal filter coefficient.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call