[Objective] The pedo-transfer functions was established through the simple and easily measurable soil properties, and the soil saturated hydraulic conductivity was obtained indirectly, in order to provide data support for soil water transport and simulation of typical agricultural small watershed of the Three Gorges reservoir area. [Methods] Using the Shipanqiu watershed of the Three Gorges reservoir area as the research object, the soil saturated hydraulic conductivity (K_s) and other basic physical and chemical properties of typical land use types (cultivated land, garden land, and grassland) were measured. In addition to correlation and principal component analysis, multiple linear regression (MLR), BP neural network (BP-ANN), and support vector machine (SVM) methods were used to construct pedo-transfer functions for the saturated hydraulic conductivity of the surface soil in the study area. Furthermore, four common pedo-transfer functions were selected to verify their applicability in this study area. [Results] The average soil K_s values were in the order of grassland>garden>cultivated land, with significant differences among different land use types. The saturated hydraulic conductivity of the soil was significantly correlated with bulk density, organic matter content, saturated water content, and soil texture. Compared with the K_s pedo-transfer functions established through multiple linear regression, BP neural network, and support vector machine, the previously used soil transfer functions model have poor prediction performance for soil saturated hydraulic conductivity in this study area. The forecast accuracy of the transfer function created using the three methods was in the order of SVM>BP-ANN>MLR, and the forecast accuracy created using principal component P₁ and P₂ as input variables was better than others. [Conclusion] The K_s values under different land use types have strong spatial variability. The pedo-transfer functions built through BP-ANN and SVM can meet the prediction requirements of K_s in this study area, and the prediction accuracy of SVM is better than that of BP-ANN.