摘要
本文提出一种基于损失最小化的SVM多类网页分类算法,该算法在多类的网页分类问题上将基于损失最小化的SVM分类算法和KNN相结合,在选择分类器顺序的问题上采用剩余样本最小错误率方法。实验表明该方法简单有效,较大地提高了SVM分类算法的准确性。
A multi-class SVM algorithm for hypertext categorization based on lost-minimization is proposed.In this algorithm,traditional KNN method is corporated into SVM while handling Multi-calss hypertext categorization tasks.The sequence of the classifiers is based on the minimization of the errors of the hold-out samples,Experimental results show the proposed algorithm can perform well with comparison to traditional SVM.
出处
《计算机应用与软件》
CSCD
北大核心
2005年第7期16-17,50,共3页
Computer Applications and Software
基金
上海市科委基础研究重点项目(02DJ14045)
南通市科委创新计划项目资助(A0020)