This study presents a dataset construction method for the cluster expansion (CE) of Au-Cu alloys. To construct the formation energy training dataset, the PBE+U ^DSCC method is employed to correct the energy levels of the Cu-3d and Au-5d orbitals, where the U and J parameters are self-consistently evaluated using the doubly screened Coulomb correction (DSCC) method. This improves the calculation accuracy of the formation energy for the Au-Cu alloys to allow the prediction of three experimentally observed stable compounds, namely Cu₃Au, CuAu, and CuAu₃. To construct the constituent strain energy training dataset, we use the PBE+U ^DSCC method to calculate the epitaxial strain energies of pure Au and Cu along different crystallographic orientations, thereby fitting the constituent strain energy of Au-Cu alloys. Based on these two datasets, a CE model corrected by constituent strain energy is established, in which the long-range elastic strain effect arising from the significant atomic size mismatch is separated from the short-range chemical interactions. Compared with the pure real-space CE model, both the fitting accuracy and the prediction accuracy of the convex hull are improved. The datasets presented in this paper are openly available at https://www.doi.org/10.57760/sciencedb.j00213.00207.