This study investigated traditional and new approaches for predicting airfield pavement responses from surface deflections measured under Heavy Weight Deflectometer (HWD) testing conducted at National Airport Pavement Testing Facility (NAPTF). In the traditional approach, pavement layer moduli were backcalculated and then pavement responses were predicted based on the multilayer elastic (MLE) theory and the finite element (FE) method. In the new approach, an Artificial Neural Network (ANN) model was developed to predict the pavement response directly from surface deflections without backcalculation. The ANN model was trained using the synthetic database that was built based on the FE simulation results using different combinations of material property, layer thickness, HWD loading magnitude, and pavement temperature. It was found that the backcalculated moduli of the asphalt surface layer were similar between MLE and FEM methods; however, discrepancies were observed for the backcalculated moduli of unbound materials. In general, the traditional approach of backcalculation and forward calculation overestimated tensile strain in asphalt layers, especially for the pavement section with a thin asphalt layer. On the other hand, the prediction accuracy of the ANN model was found better than the traditional method regarding field measurements. Further analysis of the ANN model showed that the Area Under Pavement Profile (AUPP) and Surface Curvature Index (SCI) had good correlations with critical tensile strain and shear strain in the asphalt layer, respectively.