In the paper, the superiority-inferiority in various calculations of hardenability has been compared, and the method of nonlinear regression equation was chosen to establish a mathematical model. The model was modified through the actual production of the gear steel hardenability data. Based on the model, a prediction platform was developed with computer tools to calculate the hardenability of gear steel. The hardenability prediction software can calculate random Jominy distance hardness according to chemical composition and grain size. The forecast accuracy is better than that of unmodified nonlinear regression equation and multiple linear regression equation, but equal to that of artificial neural network.