Dynamic Prediction Model of As-Built Roughness in Asphaltic Concrete Pavement Construction

Abstract

This paper develops a dynamic prediction model of a highway pavement contractor’s quality-based performance using a panel (longitudinal) data analysis. This panel data modeling uses as-built roughness measurements and pavement and contractor’s characteristics for reconstructed, replaced, and resurfaced pavement projects in Wisconsin from 1998 through 2002. Several random effects models were first developed in in-sample specification, and their modeling performances were measured by Akaike’s information criteria, which combines goodness of fit and model complexity. Out-of-sample specifications validated the developed random effects models by comparing out-of-sample forecasting accuracies. The results show that the best model has approximately a 16% mean absolute percentage error. The results finally show that asphaltic concrete pavement quality of construction can be predicted based on the contractor’s past quality-based performance and other construction parameters. Therefore, the dynamic prediction model developed in this study could be implemented in the contractor’s prequalifications required for advanced contracting methods.