The observation of soil moisture is fundamental for several climate sciences. Remote sensing had proved that it is possible to observe soil moisture from both Synthetic Aperture Radar (SAR) and SAR interferometry (InSAR) observables. This paper shows the use of machine learning regression algorithms to estimate soil moisture change using the InSAR coherence and phase and the soil type. Random Forest Regression and Extra-Tree and Bagging Regression were used. The purpose is to evaluate the improvement gain with the inclusion of “non-conventional” data such as the soil type on the estimation of soil moisture variations in time. The results point out that the inclusion of the soil type improves the estimation with coefficient of determination - R2 up to 72%.