期刊名称：International Journal of Computer Science Issues
印刷版ISSN：1694-0784
电子版ISSN：1694-0814
出版年度：2014
卷号：11
期号：3
出版社：IJCSI Press
摘要：Demand of Text Classification is increasing with the evolution of huge amount of text data available in internet, news, institutes , To make an effective text classifier we need large amount of labeled data in the form of training samples, to get labeled data is not only expensive but also time consuming, tedious task, whereas unlabelled data is easily available inexpensive. This paper proposes an algorithm that just makes use of some root words from expert followed by active search. Our algorithm also makes use of a very effective Term weighting method based on relevance factor that is used for feature representation, this text is train by BPNN. The proposed algorithm is compared on test data and on standard data 20 Newsgroup and mini Newsgroup on the basis of micro-average and macro-averaged F1 measure The Experimental results depicts the best micro averaged F1 measure of 0.95 at 2400 epochs for test data, 0.67 for 20 news group and is 0.95 for Mini Newsgroup which are comparable with the well known supervised Text classification.
关键词：Semi Supervised; text classification; Active search; term weighting method; Neural network
