Comparison of the three models used in this work. Stars and triangles correspond to the results from the RF and PRF algorithms, respectively. The results from our method are shown with open circles. We find that all the methods are consistent, indicating the robustness of the RF algorithm for the automated spectral classification of OB stars. The majority of spectral classes are predicted with a score higher than 65%, while the lower performance of some classes is either due to their inadequate representation in the initial sample or due to overlap of the EW distributions of their characteristic spectral lines with those of other classes (see Sect. 5.1).