Magnetometry is a geophysical exploration method that aims to characterize an area of interest based on its magnetic properties, being a technique highly susceptible to ferromagnetic effects. Reducing this type of artifact is challenging, making the data uninterpretable for the expert. Machine learning, along with signal processing, is a powerful tool for analyzing geophysical data. We use the wavelet transform, along with unsupervised machine learning techniques, to characterize the desired signal and distinguish it from the noisy signal; the objective then is to attenuate these effects in the data to be enhanced and further interpreted.