[Objective] The effective soil thickness of a region was rapidly and accurately obtained, and its spatial distribution and influencing factors was analyzed, in order to provide theoretical guidance for vegetation growth, soil conservation, and food security. [Methods] Taking the Xinanjiang River basin as the research area, combining field survey data, topography, lithology, climate, and other soil-forming factors, the empirical Bayesian Kriging regression prediction (EBKRP) and random forest (RF) algorithms were applied to obtain the effective soil thickness inversion results. The relationship between this data and environmental variables was also analyzed. [Results] ① The average effective soil thickness in the region ranged from 0.2 to 0.3 m. Soil thickness was higher in basin and plain areas with concentrated urban development and intensive human activity. Meanwhile, it was lower in hilly and mountainous regions. ② Based on three accuracy evaluation indicators of MAE (mean absolute error), R² (coefficient of determination), and RMSE (root mean square error), the prediction results of the RF algorithm were significantly better than those of the EBKRP algorithm. It could more effectively show the spatial heterogeneity distribution of soil thickness, improving the effect of soil thickness digital mapping. ③ The effective soil thickness estimation was strongly influenced by topography and climate variables, which accounted for 46.77% and 18.78% of the variable importance, respectively. [Conclusion] The RF algorithm could effectively invert regional effective soil thickness, overcoming the spatial heterogeneity of soil thickness, and is more accurate and has a higher resolution compared to models with limited sampling.