The key of support vector machine is to obtain a separating hyperplane, firstly to receive the initial separation hyperplane by iterative algorithm of perceptron, then to continuously rotate and translate the initial separation hyperplane, until a geometric interval up to the maximum and the complete separation of the training data set, at this moment, the separation hyperplane is approximately separation hyperplane of support vector machine, and the classification effect is the best. After the test by classification data, the result shows that this method is effective.