Using the experimental data of a wind-induced pressure coefficient, equations for the group method of data handling neural network (GMDH-NN) are developed to predict surface mean pressure coefficients (Cp¯¯¯¯$\overline{C_p}$) on the frontal surface of different C-shaped building models. Toward this objective, an extensive experiment was carried out to obtain pressure coefficients over the surfaces of the models with varying configurations, corner curvatures, and angles of incidence in a subsonic wind tunnel. The input variables include the curvature ratio (R/D$R/D$), overall side ratio (D/B$D/B$), side ratio without curvature (d/b$d/b$), height ratio (D/H$D/H$), and angle of incidence (θ$\theta$) in the radian in the GMDH-NN to develop the model equation. The performance of the GMDH-NN equation is compared with two different methods, namely, the nonlinear regression (NLR) approach through a gene expression programming (GEP) technique and a feed forward neural network (FFNN) through different statistical measures. The results indicate that the proposed GMDH-NN equation satisfactorily predicts the Cp¯¯¯¯$\overline{C_p}$ on the frontal surface with coefficients of determination (R2$R^2$) as 0.989 and 0.985 and the scatter index (SI$SI$) as 0.10 and 0.11 for training and testing data, respectively.