Comparison of Clustering Methods for Road Group Identification in FHWA Traffic Monitoring Approach: Effects on AADT Estimates

Abstract

Defining road groups is the first step in the Federal Highway Administration (FHWA) factor approach procedure for annual average daily traffic (AADT) estimation and is one of the main sources of errors in AADT estimates. This paper focuses on a comparative analysis of cluster analysis methods to identify road groups with similar traffic patterns according to different combinations of seasonal adjustment factors calculated for passenger vehicles and trucks. The aim is to highlight the differences among methods and input variables in the AADT estimation process, optimizing information commonly available to analysts. The analysis made use of traffic data from 50 automatic traffic recorder (ATR) sites in the Province of Venice, Italy. The estimation accuracy of the clustering methods was assessed and compared by considering the values of mean absolute percent error in AADT estimates. The performance of clustering methods was found to differ, depending on data sets and traffic patterns. Particularly significant for the accuracy of AADT estimates was the choice to use seasonal adjustment factors disaggregated by vehicle type as input variables.