Abstract: In this paper,we summarized the existing research about storm surge disaster,then Rough set-Support vector machine model is proposed for evaluating storm surge disaster losses through reasonable indicator system.The RS-SVM model could eliminate redundant data information,meanwhile,the model will achieve solving nonlinear problems quickly and accurately.In the later study,comparing the results about assessing storm surge losses under SVM model and RS-SVM model,we found that RS-SVM model ensured higher goodness of fit and lower error rate in the process of classification and prediction.