提出了一种基于高斯混合模型核的半监督支持向量机(SVM)分类算法.通过构造高斯混合模型核SVM分类器提供未标示样本信息,使得SVM算法在学习标示样本信息的同时,能够兼顾整个训练样本集合的聚类假设.实验部分将该算法同传统SVM算法、直推式支持向量机(TSVM)以及随机游走(RW)半监督算法进行分类性能比较,结果证明该算法在拥有较少标示样本训练的情况下分类性能也有所提高且具有较高的鲁棒性. This paper proposes a classification algorithm based on Gaussian mixture model kernel for semi-supervised support vector machine (SVM). By constructing Gaussian mixture model kernel SVM classifier to provide unlabeled sample information, SVM algorithm can learn the sample information while taking into account the whole The clustering hypothesis of training sample set is compared with the traditional SVM algorithm, TSVM and RW semi-supervised algorithm. Experimental results show that the proposed algorithm has less performance Classification performance is also improved and markedly robust with the labeled training.