Time-Series-Based Validation Methods for Microscopic Traffic Simulation Models


This paper proposes two time-series based validation methods, spectral analysis for evaluating univariate measure of effectiveness (MOE) and cross-correlation analysis for multivariate MOE. the spectral analysis quantifies and evaluates the autocorrelation after the data have been transformed into the frequency domain. the cross-correlation analysis can evaluate the stochastic relationship between two MOEs. Validation using the spectral and cross-correlation analysis is performed by constructing simultaneous confidence intervals over spectrums and cross-correlations from the field data and comparing that with the simulated spectrums and cross-correlations. the time-series based validation methods do not require the independent and identical distribution assumption, the common basis for classical statistical methods, and are more suitable for field data and simulation results as they are highly autocorrelated. the validation methods are demonstrated using two sets of field data, NGSIM data for I-80 in Emeryville, CA, and FHWA data for I-95 on Baltimore-Washington Parkway using two popular microscopic simulation models, VISSIM and AIMSUN.


Civil, Architectural and Environmental Engineering

Keywords and Phrases

Correlation Analysis; Cross Correlation; Measure of Effectiveness; Validation; Multivariate analysis; Time-series analysis; Traffic flow -- Simulation methods

Document Type

Article - Conference proceedings

Document Version


File Type





© 2008 US Department of Transportation Research and Innovative Technology Administration, All rights reserved.

Publication Date

01 Jan 2008