A speed profile, which is defined as the speed spatial distribution along a roadway facility, is a very important factor for assessing the geometric design, operations, and safety of a roadway facility. Because of the complicated lane change and deceleration maneuvers within freeway exit sections, the speed profile along a freeway exit ramp is a typical nonlinear function of ramp configuration, geometric design, traffic volume, and other factors. This study developed speed profile models for freeway exit ramps using Support Vector Regression (SVR), a novel statistical learning algorithm based on the structural risk minimization theory. Two SVR models, ε-SVR and v-SVR, were developed and tested in this study. Compared to traditional regression models, the SVR models can improve the prediction accuracy in terms of Mean Absolute Percentage Error (MAPE) and Mean Square Error (MSE) for predicting the speed profile along a freeway exit ramp.