Development of an alternative approach to estimate vehicle miles traveled using a classification procedure

Abstract

In this study a Classification and Regression Tree (CART) classification system method was developed as an alternative to the HPMS conventional method of VMT estimation. Louisiana Highway Performance Monitoring System (HPMS) data from Federal Highway Administration (FHWA) and Louisiana highway network data from Louisiana Department of Transportation (LADOTD) were used to carry out the analysis. The CART classification was derived based on link attributes that effectively distinguished the homogeneous traffic volume sections. VMT estimation on individual links as well for the different systems was calculated using the CART classification system and the values were assigned on each link. The default MOBILE6 VMT values were also assigned on each link. The entire procedure was carried out using GIS Developer’s kit in TransCAD. The whole procedure was automated by creating a customized add-in program to TransCAD. The resulted AADT estimates from the CART classification system equation were compared with the HPMS conventional method estimates and with the observed values on the HPMS sample sections. This comparison was done graphically and statistically by the paired sample T-test. The results indicated that the HPMS conventional method performs better in estimating AADT than the CART classification system method. The study also included the demonstration of the CART classification system add-in the East Baton Rouge non-attainment area. The developed add-in can also be used on any other traffic count data for any region, state or country. The GIS program developed during this study provides a framework, which can be built upon further and shared with other researchers in future.