Therefore a study on the predicting model with non stationary time series is made.

By using the proposed segmental feature dependent non stationary time series model, the new model not only achieves the modeling of correlation between different scale features but also implicitly models the correlation among neighboring frames in frame scale via parametric mean trajectory function.

This paper introduced a method of wavelet based analysis and forecasting for non stationary time series.

Study on the methods for modeling and forecasting gyro s drift performance based on non-stationary time series;

Non-stationary time series prediction based on wavelet analysis and AR-LSSVM;

To build an appropriate mathematic model of the random drift of flexible gyro,two methods are discussed,using non-stationary time series analysis and using stationary time series analysis based on wavelet discomposition.

The moving average model-based nonstationary time series method is used in the model distinction, parameter estimation, and new information prediction in this paper.

? A state space approach for the modeling of nonstationary time series is presented.

This method is used to build forecasting models on the ideal time series from 33-mode Lorenz system,and especially the prediction on nonstationary time series are tested and analyzed.

The paper suggests the forecasting model about objects that have nonlinear time sequence by using neural network model, and verifies in through an example.

The complexity measure is an important dynamic index to describe nonlinear time sequence.

Correlation dimension is an important parameter to measure a nonlinear time sequence quantitatively,and it is widely used to analyze harmonic component of power system and biomedical signal.