This study demonstrates the use of neural networks in predicting hydraulic flow units depends on integrating well logs, routine core data, and minerals analysis by using x-ray diffraction analysis and scanning electron microscopy, to overcome the reservoir heterogeneity in Obaiyed field, Western Desert, Egypt. For this purpose, well-log and core data from five wells (D10, D13, OBA J14-2, OBA 2-3, and D7A) in Obaiyed gas field (Jurassic Lower Safa reservoir [Khatatba Formation]) were used. Electrofacies model indexed and probabilized self-organizing map was built using suite of well logs analyzed into their principal components as the model input along with the rock typing data derived from the cores as the model learning. This model was built using an artificial neural network on which the learning and the validation were performed on cored wells. Connect x-ray diffraction analysis data result and that from the electrofacies model to assess the clay occurrence in each hydraulic flow unit, which showed a very high impact on the reservoir permeability. Using cored well data covering various reservoir properties, the reservoir can be divided into different hydraulic flow units.