Papers by Author: Hong Zhang

Abstract: Rescaled range analysis (R/S) method is a scaling method commonly used for detecting the long-range correlations in many time series. The aim of this paper is to show that, using the rescaled range analysis on sunspot time series, how the threshold values q affects the correlations of the return intervals for events above a certain threshold q. We find that both the original records and the return intervals are long-range correlated.

Abstract: Rescaled range (R/S) analysis has regained in popularity as a robust tool for the examination of long-term dependence. This paper modified the R/S method to determine the periodicity. We firstly apply the modified method to artificial time series for investigating the validity of the new method. Then the proposed technique is employed to research periodicity of the sunspot time series and the results indicate the robustness of the modified method.

Abstract: In this paper, we analyze the stock of Nanjing Panda Electronics Co Ltd for the 44-year period, from May 2, 1996, to October 9, 2009, a total of 3200 trading days. Using the Box-counting dimension method, we find that the financial data have different power law exponents in the plot for the number of box and diameter of box, which indicates the multifractality exist in the time series. In order to investigate the latent properties in the data, the width and maximum of the singular spectrum are calculated. The results show the strong degree of multifractality in the time series.

Abstract: In this paper, we analyze the stock of Tsingtao Brewery Co Ltd for the 8-year period, from July 31, 2001, to September 11, 2009, a total of 2003 trading days. Using the False Nearest Neighbors method, we obtain the embedding dimension m in the k-nearest neighbour Algorithm. In order to investigate the validity of this method, we apply the modified method to the daily adjusted opening values of the Tsingtao Brewery Co Ltd. We find that the prediction of experimental results is more accurate than traditional methods.

Abstract: We propose a new method called the multi-dependent Hurst exponent to investigate the correlation properties of the nonstationary time series. The method is validated with the artificial series including both short-range correlated data and long-range correlated data. The results indicate that the multi-dependent Hurst exponents fluctuate around the a-priori known correlation exponent H. Application to traffic time series is also presented, and comparison is made between the artificial time series and traffic time series.