[Objetcive] An evaluation model for the soil and water conservation function of vegetation at community scale was explored in order to provide a reference for soil and water conservation and ecological construction of vegetation in the Dabie Mountains. [Methods] Based on existing research, through the bibliometric method and expert experience method, the primary selection index for soil and water conservation function of vegetation was established, including 30 items characterizing vegetation type, vegetation structure, terrain, and so on. The measured values of the indicators were obtained through field investigation, and the indicators were analyzed and measured by the random forest method and the decision tree method. A multiple stepwise regression relationship between the measured values of the indicators and soil erosion was established to construct an evaluation model for soil and water conservation function of vegetation at community scale. [Results] A global regression model based on a random forest (R²=0.71) was better than the local regression model based on a decision tree (R²=0.69). Based on the global regression model of random forest, an evaluation model which included slope, forest age, shrub coverage, grass coverage, and litter thickness was constructed. The soil conservation functions of vegetation in the study area followed the order of protection forest (0.705)> timber forest> (0.529) economic forest (0.513)> shrub-grass (0.457). [Conclusion] In the evaluation of vegetation soil and water conservation function at community scale, the model constructed by random forest method has strong applicability, operability and is easy to grasp.