Similarity Analysis on Nonstationary Time Series

In order to study similarity between nonstationary time series, in this paper, we analyze clusters of size series of cocoon filament (SSCF) which are considered be nonstationary in terms of both mean and auto-covariance. We obtain model parameters of these clusters, such as coefficients of regression equation of deterministic components and parameters of time varying parameter autoregressive (TVPAR) model of stochastic components. By means of stochastic analysis, similarity measuring methods for nonstationary time series are proposed. By use of these methods, cluster analysis on SSCF is carried out. The classifying method proposed in this paper is simpler and has more precision. The results show that the similarity measuring method is feasible.