In order to enhance transformer fault classification accuracy based on dissolved gas analysis (DGA) and overcome the limitations of single subspace, a new transformer fault multilayer diagnosis model based on dual-space feature extraction algorithm was proposed. Firstly, a DGA test sample was projected to a subspace to realize feature extraction in order to reduce the dependence of the modeling accuracy on kernel function and parameters and take advantage of stronger robustness and higher precision. Multiple-kernel support vector machine (MKSVM) was used as the classifier to predict the class label. The test sample was classified into difficult class one or easy class one according to the predicted result, and the class label of the easy one was identified in the subspace directly. As to the difficult one, the test sample was re-projected to another subspace where multiple-kernel support vector machine was used to predict. The class label of the difficult was identified to integrate two predicted results. Therefore, MKSVM of two class problem based on dual-space feature extraction algorithm was achieved. Finally, a multilayer diagnosis model was established according to the fault characteristic of transformer. The diagnosis experiment has shown that the model has a higher diagnosis rate, which proves its effectiveness and usefulness.