点击次数：

论文类型：期刊论文

发表时间：2009-06-01

发表刊物：Journal of Computational Information Systems

收录刊物：EI、Scopus

卷号：5

期号：3

页面范围：1373-1378

ISSN号：15539105

摘要：Self-Training algorithm is a semi-supervised classification algorithm which through repeated training with the labeled data to get a enlarged labeled data set and improve the classification accuracy meanwhile. Since the initial labeled data set in Self-Training algorithm may be small, a considerable number of data are mislabeled in the training process is unavoidable. A nearest neighbor rule based data editing technique is introduced, which extends traditional self-training algorithm by new methods of identifying and removing the mislabeled data, so that it can reduce the mislabeled data and improve the classification accuracy. The data sets used in experiments are all from the UCI machine repository. The classification effect is improved in different levels through contrast. The experimental results show that the introduction of the data editing technique is beneficial for improving the classification effect of Self-Training. 1553-9105/ Copyright ? 2009 Binary Information Press.