Search

Complex network theories may be a powerful tool in time series analysis. But when facing high frequency data, current method of building network is highly inefficient. Thus, this paper proposed a symbolic method for time series compression and built network. This method used least square to estimate the slope of each segments and then extracted local features from time series. Then it determined node's adjacency through a dictionary. Simulation studies show the scaling exponent, cluster coefficient of the network are highly correlated with Hurst exponent and can exactly capture complex properties of original series. [ABSTRACT FROM AUTHOR]