The dual support vector machine(Dual-SVM) is proposed to promote the speed of establishing model and
diagnosing. The diagnosis model is established by two SVM training processes. In the first process, the approximate
classifying hyperplane is directly obtained by the two classes centers and the distribution of the samples on the connecting
direction of centers. In the second fuzzy SVM process, the boundary samples are selected, fuzzy memberships are calculated,
and the diagnosis model is established. The experiments on DARPA data-sets show that the Dual-SVM can get higher training
speed and more simplified model compared to SVM.