Comparison between the performance of GP and SVM with either the radial or the normalized polynomial (NP) kernel. Ten models were calculated for each dataset and for each combination of modeling technique and kernel, thus resulting in a total of 60 models. The performance of GP and SVM was assessed by kernel for the three datasets. Given that the distributions of RMSEPext and values were normally distributed, a two-tailed t-test of independent samples was applied to statistically evaluate their differences. These analyses let us conclude that SVM and GP perform on par for the modeling of the three datasets considered in this study.