 This study aimed to improve soil salinity prediction by combining satellite data and machine learning algorithms. It evaluated the performance of various machine learning models overcoming the limitations of conventional techniques and optimizing the variable input combinations. The results showed that the best model was a combination of the BPNN feature selection method with the random forest regression learner RFRL which achieved a root mean square error RMSE of 0.000246. This model can be used to provide farmers with more accurate information about soil salinity levels allowing them to make informed decisions regarding planting procedures and improving the sustainability of their lands. This article was authored by Syed A. Mohamed, Mohamed M. Mevely, Mohamed Armitwali and others.