为解决支持向量机在分类识别前需要利用已知训练集进行训练的问题，本文提出了一种基于k均值的对无标识数据进行分类的支持向量机分类算法。首先利用k均值算法将未知数据划分成某个数量的子集，然后对新数据进行支持向量机训练得到决策边界与支持矢量，最后对无标识数据进行分类。模拟结果表明：训练时消耗的CPU时间为1.8280秒，支持向量个数为60时，分类错误率小于2％。 In order to solve the problem that SVM needs to train with known training set before classification identification, this paper proposes a new SVM classification algorithm based on k-means to classify unlabeled data. Firstly, the unknown data is divided into a certain number of subsets by using the k-means algorithm. Then, the new data is trained by SVM to get the decision boundary and support vector. Finally, the unclassified data is classified. Simulation results show that the CPU time spent on training is 1.8280 seconds, and the number of support vectors is 60, the classification error rate is less than 2%.