This is an interpretation technique for numerical weather prediction (NWP) products of GFS. In this paper the support vector machine (SVM) was employed to forecast the intensity of offshore and landing tropical cyclones (TC). According to the TC intensity, meteorological environment and topographical factors, we designed related factors and built a forecasting model which was used to forecast the intensities of TC in 12, 24, 36, 48, 60 and 72 hours. The predicting precision by SVM is closer to the forecast by the National Meteorological Centre of CMA and all predictions are superior to that by the method of climate persistence. The trend prediction of TC has obvious advantages and can improve forecast precision by 7%-12%. All these indicate that SVM is a better method and can be applied to TC intensity forecast in operations.