For modelling the flow transport in unsaturated conditions, we can use hydraulic properties which are expensive and time-consuming to be obtained directly because of high variability and complexity of soil systems. Few studies have been done about pedotransfer functions (PTFs) in smectitic soils. Moreover, the utility of fractal parameters in the prediction of soil water retention curve (SWRC) have not been investigated in these soils. In this study, PTFs have been made for estimating the parameters of van Genuchten (VG) and Dexter models by regression and artificial neural networks methods. Therefore, 69 soil samples were collected from Guilan Province, Iran. Fractal and non-fractal models were fitted to the particle size distribution (PSD) and micro-aggregate size distribution (ASD) and their parameters were calculated. To create PTFs, the parameters of PSD and ASD models were used as estimators. The comparison of the results of the two models of Dexter and VG shows the priority of Dexter model for the purpose of testing of smectitic soils. The results showed the superiority of Fredlund et al. PSD model parameters and fractal parameters of ASD, in the estimation of Dexter and VG SWRC models, respectively. This outcome may be related to the higher accuracy of Fredlund et al. PSD model in the description of the PSD data in the clayey soils. However, the higher number of parameters in comparison to the number of fractal model parameters may be another reason.