On the base of current researches on multiclass classification with support vector machine, an incomplete binary tree SVM multi class classification algorithm based on hypersphere is proposed. The algorithm adopts hypersphere SVM algorithm to calculate the distribution of each sample groups. Then, the distance formula is used to calculate the distance among the sample classes. According to the principle that the class which can be separated easiest must be split first, the algorithm designs binary tree to improve the classification accuracy. Compared with many classification methods, the effectiveness of the algorithm is verified by simulation experiments.