Abstract: Aiming at the problem of materials corrosion prediction in the natural soil environment, hierarchical linear modeling was used to study the corrosion rules of materials. The hierarchical linear model is a new statistical technique for hierarchical data analysis. Different models for every region can be established to fit regional soil corrosion characteristics and solve the problem of small samples in a specific region, which makes modeling assumptions more consistent with actual situations. A hierarchical linear corrosion model was built with carbon steel corrosion data in soils being the object of study. The test demonstrates that the model can precisely describe and predict the corrosion rate of carbon steel in soils, and it is superior to exponential smoothing algorithm and differential autoregressive integrated moving average (ARIMA) algorithm.