Abstract: Based on the principle of support vector machine (SVM) and twoclass or multiclass classifier,a model of evaluating flood disaster loss was established in consideration of the character of many factors which are related to the evaluation and the small samples in practice,and some problems which are difficult or impossible to be resolved in present model, such as overfitting,the local minimization and so on.The model was tested in a case which represented real flood disasters happened in some areas in China from 1989 to 1990.It is proved that the model is as useful as the model of artificial neural network (ANN) and other traditional methods,the time for calculating this model is much less than normal models.So the model established in this paper will be much more applicable in the future in the evaluation of flood disaster loss because of minimization of structure risks and its advantage in solving limited sample, nonliner and high dimensions recognization. 〖

Key words: support vector machine, evaluation model, flood disaster loss〖