基于Transfer-SVM多标签文本分类算法研究
|
摘要
传统的支持向量机分类模型只有在利用大量已标注数据进行训练才能获得较高精度。在实际应用中,多标签数据相对于传统单标签数据更具有价值,但多标签数据中含有大量冗余数据,获取大量多标签数据难度非常大。文章提出一种基于迁移学习的分类算法,利用目标数据域和源数据域的相关性,从源数据域中选取对分类超平面起关键作用的支持向量和目标数据域,一起训练分类模型以提高分类精度。
The traditional support vector machine classification model can obtain high precision only if it is trained by using a large amount of labeled data. In practical application, multi-label data is more valuable than traditional single-label data, but multi-label data contains a lot of redundant data. So it is very difficult to obtain a large number of multi-label data. In this paper, a classification algorithm based on migration learning is proposed, which uses the correlation between the target data domain and the source data domain to select the support vector and the target data domain which play a key role in the classification hyperplane from the source data domain to train the classification model to improve the classification accuracy.
The traditional support vector machine classification model can obtain high precision only if it is trained by using a large amount of labeled data. In practical application, multi-label data is more valuable than traditional single-label data, but multi-label data contains a lot of redundant data. So it is very difficult to obtain a large number of multi-label data. In this paper, a classification algorithm based on migration learning is proposed, which uses the correlation between the target data domain and the source data domain to select the support vector and the target data domain which play a key role in the classification hyperplane from the source data domain to train the classification model to improve the classification accuracy.
引文
[1]JIANG S,PANG G,WU M.An improved K-nearest-neighbor algorithm for text categorization[J].Expert Systems with Applications,2012(1):1503-1509.
[2]SEBASTINAI F.Machine learning in automated text categorization[J].Association for Computing Machinery Surveys,2002(1):1-47.
[3]YANG J,YAN R,HAUPTMANN A G.Cross-domain video concept detection using adaptive SVMs[C].Augsburg:the 15th International Conference on Multimedia,2007.
[4]CHIH C C,CHIHJEN L.LIBSVM:a library for support vector machine,2001[EB/OL].(2018-07-15)[2019-05-10].http://www.csie.ntu edu.tw/~cjlin/libsvm.
[2]SEBASTINAI F.Machine learning in automated text categorization[J].Association for Computing Machinery Surveys,2002(1):1-47.
[3]YANG J,YAN R,HAUPTMANN A G.Cross-domain video concept detection using adaptive SVMs[C].Augsburg:the 15th International Conference on Multimedia,2007.
[4]CHIH C C,CHIHJEN L.LIBSVM:a library for support vector machine,2001[EB/OL].(2018-07-15)[2019-05-10].http://www.csie.ntu edu.tw/~cjlin/libsvm.