This study was conducted to estimate the dynamic characteristics of clean sand under low strains using fuzzy expert systems and neural network approximations. A series of resonant column tests were conducted on clean sand specimens to create a large database. The effects of various factors, such as effective pressure, saturation, void ratio and shear strain levels, were simulated using fuzzy expert systems and neural networks. The neuro-fuzzy inference method was employed to predict the initial shear modulus of clean sand samples as a substitute for time-consuming laboratory testing. Additionally, the maximum shear modulus results were compared with the existing empirical relationships. From these observations, it can be observed that certain relationships significantly underestimate the initial shear modulus. Simple empirical relationships to estimate the initial shear modulus were formulated. It is concluded that neuro-fuzzy-based models provide useful guidelines for the preliminary estimation of the dynamic shear modulus for clean sand soils.