Support vector machine (SVM) is a novel machine learning based on statistical learning theory. It is powerful for the problem with small samples, nonlinear and high dimension. Multi-class support vector machine is extended for multi-class classification based on traditional SVM which is a classifier only for binary classification. A model of transformer fault diagnosis based on Multi-class SVM is present in this paper. It uses the grid search method based on cross-validation to determine the model parameters. Taking into account the compactness characteristics of DGA data, the fuzzy C-means (FCM) clustering method is adopted to pre-select samples achieved. Compared with the model based on layered combined binary SVM, Multi-class SVM classification is conveniences in model construction and parameters selection. Practical analysis shows that this model has good classification result and extension ability.