Abstract: Soil samples were collected from 101 different geographical locations in Beijing, and nine physical and chemical parameters of the soil samples were tested in the laboratory to obtain the distribution range of soil parameters in Beijing. Got through the analysis of the machine learning in the Beijing area soil corrosive key influencing factors for the corrosion potential, soil moisture content and soil resistivity, and the corrosion rate prediction model was built based on random forest algorithm, the predicted values and the roughly tallying with the actual values, the average absolute error less than 5%, the smaller values of the prediction error, show that the prediction model has higher prediction accuracy. In order to deeply explore the relationship between three key soil parameters and soil corrosion rate, the established corrosion rate prediction model was used to take three key parameters such as self-corrosion potential, soil resistivity and soil water content as the input and corrosion rate as the output. The prediction results showed that: When the self-corrosion potential is between -0.5 and -0.7VSCE, the water content is between 15% and 22%, and the soil resistivity is between 30 and 80 Ω·m, the soil corrosion rate is higher than 0.1mm/y. The results provide a relatively simple method for the evaluation of soil corrosion in Beijing.