Computer Engineering and Applications ›› 2013, Vol. 49 ›› Issue (14): 144-146.

Previous Articles     Next Articles

k means clustering based transductive support vector machine algorithm

WANG Limei1, LI Jinfeng1, YUE Qi2   

  1. 1.Institute of Engineering, Mudanjiang Normal University, Mudanjiang, Heilongjiang 157011, China
    2.College of Engineering and Technology, Northeast Forestry University, Harbin 150040, China
  • Online:2013-07-15 Published:2013-07-31

基于k均值聚类的直推式支持向量机学习算法

王立梅1,李金凤1,岳  琪2   

  1. 1.牡丹江师范学院 工学院,黑龙江 牡丹江 157011
    2.东北林业大学 信息与计算机工程学院,哈尔滨 150040

Abstract: As transductive support vector machine runs slowly, this paper proposes a k means clustering based transductive support vector machine algorithm. The algorithm utilizes k means clustering to divide the unlabeled samples into several clusters, labels them with the same class, makes transductive inference on the mixed data set composed by both labeled and unlabeled samples. As TSVMKMC algorithm reduces the size of the state space effectively, the running speed is improved largely. The experimental results show that the algorithm can achieve good classification accuracy with faster speed.

Key words: transductive inference, support vector machine, k means clustering, unlabeled samples

摘要: 针对直推式支持向量机(TSVM)学习模型求解难度大的问题,提出了一种基于k均值聚类的直推式支持向量机学习算法——TSVMKMC。该算法利用k均值聚类算法,将无标签样本分为若干簇,对每一簇样本赋予相同的类别标签,将无标签样本和有标签样本合并进行直推式学习。由于TSVMKMC算法有效地降低了状态空间的规模,因此运行速度较传统算法有了很大的提高。实验结果表明,TSVMSC算法能够以较快的速度达到较高的分类准确率。

关键词: 直推式学习, 支持向量机, k均值聚类, 无标签样本