In machine learning, support vector machine has become a popular method in classification and regression. On the other hand, conformal inference has also shown good properties in regression prediction. In this research, we compare these two approaches in precision of prediction of future outcomes in the regression settings based on training data. Even though these methods were proposed to be distribution independent, however, the level of performance in prediction can actually be affected by the underline unknown data distribution. An extensive simulation is conducted for the comparison and a data set from a recent clinical trial is used to illustrate the proposed approach.