The scarcity of available and suitable construction sites in urban centers has led to greater use of problem areas, where the carrying capacity of underlying deposits is very low. Reinforcing these problem soils with granular filler layers is one of the most widely used soil improvement techniques. The problem behavior of the soil can be improved by totally or partially replacing the inadequate soils with layers of compacted granular filler. The study presented here describes the use of artificial neural networks (RNAs) and the multi-linear regression model (MLR) to predict the carrying capacity of circular surface bases supported by layers of compacted granular filler over natural clay soil. The data used in the operation of the network models have been obtained from an extensive series of field tests, including large-scale base diameters. Field tests were performed using seven different base diameters up to 0.90 m and three different granular filler layer thicknesses. The results indicate that the use of granular filler layers on the natural clay soil has a considerable effect on the load bearing characteristics and that the ANN model serves as a simple and reliable tool for predicting the load capacity of circular footings on stabilized natural clay soil.