The creep analysis of concrete is of great importance in stress calculation and deformation analysis of concrete structures. Since many factors affect concrete creep, the creep function is highly nonlinear, making it difficult to obtain satisfactory results using the traditional regression method. The gene expression programming (GEP) method, which combines the advantages of GA and gene programming, shows great promise and capability of data mining in nonlinear fitting. Based on theoretical analysis of GEP, the data mining model of the creep function was established by using GEP. Examples indicate this model is more accurate than functions obtained by conventional methods.