In this paper results of an experimental investigation on the bearing stiffness (K f ) of Ultra-High Performance Concrete (UHPC) under dowel bars are summarized. The effect of concrete strength, bar diameter, and location of the bar in concrete were investigated. By considering these parameters as input variables, several linear and nonlinear regressions and also Support Vector Regressions (SVRs) by incorporating different kernels are constructed, trained and tested to predict the K f of UHPC. Comparing the results show that in the regression models, quadratic polynomial is more feasible in predicting the K f of UHPC than other proposed functions and among the various kernels, SVR with radial basis function (RBF) kernel exhibits better results than other kernels.