Quantification of Root-Zone Soil Moisture (RZSM) is crucial for agricultural applications. It impacts processes like vegetation transpiration and water percolation. The surface soil moisture (SSM) can be assessed through active and passive microwave remote sensing, but no current sensor enables direct retrieval of RZSM. Spatial maps of RZSM can be retrieved via proxy observations (vegetation stress, water storage change, surface soil moisture) or from land surface model predictions. Recently, more interest has risen in the use of data-driven methods to predict RZSM. In this study, we investigated the use of physical-process based features in the context of Artificial Neural Networks (ANN). We integrated the infiltration process information into an ANN model through the use of the recursive exponential filter. We also used a remote sensing-based evaporative efficiency as an input feature. It is important to note that these two processes depend on surface soil moisture which can be assessed through remote sensing. The impact of the use of geophysical variables was also assessed through the use of surface soil temperature and Normalized Difference Vegetation Index (NDVI). At each step of the study, the ANN models were trained using either only in-situ surface soil moisture data provided by the International Soil Moisture Network (ISMN) or an additional geophysical or processbased feature. The results show that the use of more features in addition to SSM information improves the prediction accuracy in specific cases when compared to an ANN model that predicts RZSM based on only SSM. The ability of the developed models to predict RZSM over larger areas will be assessed in the future.