This work aims to implement and analyze the nonparametric kernel ridge regression (KR) proposal developed in Filipovic, Pelger and Ye (2022) for the construction of the term structure of interest rates from Brazilian federal government bonds. Specifically, we investigate how this model compares with other parametric models in terms of usability, accuracy and robustness. We build the historical yield curves not only for the KR model, using Filipovic et al (2022) own routines, but also for the ANBIMA reference model based on Svensson (1994). As in Gilli, Große and Schumann (2010), we implement the parametric optimization of the factor model of Svensson (1994) using the genetic algorithm of Differential Evolution of Storn and Price (1997).