In order to enhance the generalization performance and prediction accuracy of Least Square Support Vector Machines for the wind pressure on the building surface, the method of increasing the number of input points has been proposed in this paper. Through increasing the number of input points, more information can be providedfor least squares support vector machines, which in turn optimizes the performance of the forecasting model of least squares support vector machines. Then interpolation prediction and extrapolation prediction of measured wind pressure at different lengths are carried out respectively. Experiment results show thatwhen interpolation prediction and extrapolation prediction are used, the performance of wind pressure prediction can be improved through the increase of the number of input points, but with interpolation prediction,the performance is better than that of extrapolation prediction. When the number of training sets increases, increasing the number of input points has little effect on prediction performance.