This study proposed a yield curvature prediction model considering axial compression ratio with exponential function which is improved on the basis of the specification effective yield curvature prediction model. Parametric moment–curvature curves approach was used to verify that the yield curvature is greatly affected by the section size, the yield strength of longitudinal steel bar and the axial compression ratio. On the basis, the yield curvature under different levels was obtained by Xtract and parametric moment–curvature curves approach, the result shows that with the increase of axial compression ratio, the yield curvature also increases, which is roughly a linear relationship. Subsequently, combined with the specification and considering the influence of axial compression ratio, a new yield curvature prediction model is proposed based on a massive sample space obtained by parametric moment–curvature curves approach. Moreover, by comparing with the experimental database of PEER, the accuracy of new prediction model is verified, and the result shows that the new prediction model is suitable for estimating yield curvature of square section columns.