Support vector machine is widely used in pattern recognition problems such as the portrait recognition and the text classification recognition. It can effectively solve some classification problems in real life. In this paper, a classification algorithm based on Seeded-Kmeans and SVM （SK-SVM） is proposed for the semi-supervised two classification problem. The Seeded-Kmeans algorithm is used to process the unlabeled points to obtain initial labels. Then, effective label points are selected and added to the existing labeled points to form a new labeled training set. Finally, SVM is combined to classify the unlabeled points.