This paper presents a fast intra mode decision solution for the VVC standard using machine learning. The idea is to reorder the evaluation of modes performed by the Rate-Distortion Optimization (RDO) process according to the modes occurrence rate. Based on the new evaluation order, three Decision Tree models are trained to skip the modes that are less likely to be chosen. The results show that the proposed solution achieves time savings of up to 15.57% with coding efficiency degradation of only 0.41% on average. When compared with related works the proposed solution shows competitive results.