Traditional pavement design relies on computational deterioration models that were trained using relatively small datasets and solid mechanic theory. This paper analyzes full-depth reclamation (FDR), a pavement recycling rehabilitation alternative that is not modeled effectively in current design models. A random forest algorithm is calibrated based on empirical data collected from 10 FDR sites in Colorado. The resulting model yields significant improvements compared to the mechanistic-empirical model currently used in pavement design, with a reduction in RMSE of 86.7%. The study found that precipitation and freezing index have the greatest impact on rutting depth at the end of the pavement design life, with changes in traffic having a significantly lower impact on rutting. This study provides a framework for utilizing large amounts of condition, climatic, and traffic data to model pavement deterioration.