A wind tunnel test is conducted in this study on the scaled model of the Guangzhou International Sports Arena (GISA). Simultaneous pressure measurements are conducted in a simulated suburban boundary layer flow field. A numerical simulation approach using Fuzzy Neural Networks (FNNs) is developed for the predictions of wind-induced pressure time series at roof locations which are not covered in the wind tunnel measurements. On the other hand, the wind-induced response of the roof are presented and discussed, which are directly calculated by the Complete Quadratic Combination (CQC) approach. Furthermore, the correlations between the background and resonant response components are discussed in detail, and the results show that neglecting the correlations between the two components would result in considerable error in the response estimation. Finally, the Equivalent Static Wind Load (ESWL) approach is used to estimate the wind-induced responses of the roof, which are compared with those obtained from the CQC approach to examine the effectiveness of the proposed ESWL approach in the design and analysis of large-span roof structures. It is shown through the example that the FNN and ESWL approaches can successfully predict the wind-induced pressures and responses respectively. [ABSTRACT FROM AUTHOR]