摘要
SVM算法只使用已归类的数据训练分类器,而EM算法用少量已归类数据,结合大量的未归类数据来训练分类器,在减少已归类数据的同时保证了分类器的精度。本文基于EM算法的思想,根据SVM文本分类模型,提出一种新的迭代SVM文本分类算法。实验结果表明,迭代SVM算法分类精度高于传统的SVM文本分类算法,具有较好的性能。
SVM algorithm use only labeled data to build a model for the classifier, then the EM_BAYES classifier trains the model by a smaller number of labeled data augment with a large number of unlabeled data and ensures the accuracy. In this paper, we present a text classifier, iterative SVM, which combines the EM algorithm with the SVM model. The experiment results show that the iterative SVM classifier achieves better performance than the SVM text classifier.
出处
《衡阳师范学院学报》
2006年第3期97-99,共3页
Journal of Hengyang Normal University
基金
湖南工程学院资助项目(K04022)
关键词
SVM算法
EM算法
文本分类
半监督学习
SVM algorithm
EM algorithm
text classification
semi supervise learning
