The Analysis And Prediction Of Several Groups Of Non-linear Dynamic Data

Chaos is an irregular behavior which is a ubiquitous in nature and seemingly random generated by the nonlinear deterministic dynamical systems. Chaotic time series analysis and prediction is an important problem. In this dissertation we studied the calculation law of chaotic characteristics of different data, employed the quantitative methods to analyze the chaotic characteristics of the actual data, and studied the predictability of the data. Exploring the chaotic characteristics of the actual data, it provided a better support for the further study of the description of chaotic properties and chaotic prediction. We employed the autocorrelation function method to estimate the embedded delay and employed the false nearest neighbor method to estimate the embedding dimension. We analyzed chaotic characteristics based on Phase Space Reconstruction theory. We judged whether the dy-namical system is chaotic by the largest Lyapunov exponent. We analyzed and predicted Hong Kong weather data, the U.S. Nasdaq index data, China Southern Airlines company stock data, Vanke quote data. Firstly,we employed the autocorrelation function method to obtain embedded delays and employed the false nearest neighbor method to obtain embedding dimensions. Then we calculated their largest Lyapunov exponents employing Tisean software, and determined whether they were chaotic. At last we predicted them by RBF neural network model. We also wrote programs to analyze and predict some data based on Matlab software.