The application of machine learning in geophysics has increased dramatically in the last decade, with the quality of its results varying depending on the type of seismic problem in focus and the computational method used. Deep learning methods are achieving impressive results in this area, but we note that there is still a lack of certainty about whether classical machine learning methods can provide similar results. In the present work, the aim was to try to fill part of this gap by comparing a well-known deep learning method with a non-deep learning based machine learning method for direct wave propagation and seismic inversion problems for layered 2D models. horizontal. Both methods are evaluated in different scenarios, but under similar conditions, to be able to understand the effect of parameter configuration on their results. The dataset has 20,000 samples, each consisting of three vectors: a velocity vector with 236 values (representing a vertical profile of a randomly generated 2D layered model), a reflective vector with 600 values obtained directly from the velocity vector, and the associated vector seismogram with 11 traces containing 600 values each. The overall results show that WaveNet achieves a lower MSE between predicted and correct results than XGBoost results. A challenge that has not yet been faced is that WaveNet can train well on a GPU, but it was not possible to do the same with XGBoost, due to the amount of data to be processed.