A Combination of Differential Evolution and Support Vector Machine for Rainstorm Forecast

This study employed a DE-SVM model that hybridized the differential evolution (DE) and support vector machines (SVM) to improve the classification accuracy for rainstorm forecasting. This optimization mechanism combined the DE to optimize the SVM parameter setting. Based on the European Centre for Medium-Range Weather Forecasts (ECMWF), Japan and T213 precipitation data from 2003 to 2006, using DE-SVM, the 24 hour´s storm models for 5 sub-regions in Hubei province were created, which have been used in the real-time running work from May to July in 2007. The results have shown the forecasting ability and reference value of the SVM method.