Fitting and Characteristics Analysis of Travel-Time Fluctuations on an Urban Road Network

Abstract

Detailed knowledge about the times at which links in a network suffer travel time instability is of great significance for route decision making and traffic management. With a data set of 1,170 links in a large-scale urban road network, this paper introduces advanced algorithms to fit the travel time volatility (TTV) and analyze its characteristics. TTV is prone to clustering, and its distribution exhibits leptokurtosis and a heavy tail. These characteristics suggest iterative cumulative sums of squares (ICSS) and the autoregressive conditional heteroskedasticity family models (ARCHs) are applicable to analyze TTV. ICSS finds the structural change points to locate the time and determine the intensity of the travel time variances fluctuate. ARCHs are employed to fit the TTV series, and the best model is determined by comparison. ICSS and ARCHs are combined to improve the fitting model. With TTV, TTR is analyzed from a dynamic and temporal perspective at intervalwise and linkwise levels.